Minority Report
American Politics and Political Economy A series edited by Benjamin I. Page
Minority Report evaluat...
58 downloads
746 Views
1MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Minority Report
American Politics and Political Economy A series edited by Benjamin I. Page
Minority Report evaluating political equality in america
john d. griffin and brian newman
The University of Chicago Press Chicago and London
john d. griffin is assistant professor of political science at the University of Notre Dame. brian newman is assistant professor of political science at Pepperdine University. The University of Chicago Press, Chicago 60637 The University of Chicago Press, Ltd., London © 2008 by The University of Chicago All rights reserved. Published 2008 Printed in the United States of America 17 16 15 14 13 12 11 10 09 08 isbn-13: 978-0-226-30867-8 (cloth) isbn-13: 978-0-226-30868-5 (paper) isbn-10: 0-226-30867-7 (cloth) isbn-10: 0-226-30868-5 (paper)
1 2 3 4 5
Library of Congress Cataloging-in-Publication Data Griffin, John David, 1968– Minority report : evaluating political equality in America / John D. Griffin and Brian Newman. p. cm.—(American politics and political economy) Includes bibliographical references and index. isbn-13: 978-0-226-30867-8 (cloth : alk. paper) isbn-10: 0-226-30867-7 (cloth : alk. paper) isbn-13: 978-0-226-30868-5 (pbk. : alk. paper) isbn-10: 0-226-30868-5 (pbk. : alk. paper) 1. Political participation—United States. 2. Minorities—United States—Political activity. 3. Representative government and representation—United States. 4. Proportional representation—United States. 5. Pluralism (Social sciences)—United States. 6. United States—Politics and government. 7. United States—Race relations. 8. United States—Social conditions. I. Newman, Brian, 1975– II. Title. J K 1764.G743 2008 323'.042080973—dc22 2007050136 ∞ The paper used in this publication meets the minimum requirements of the American National Standard for Information Sciences— Permanence of Paper for Printed Library Materials, ansi z39.48-1992.
Contents List of Figures | vii List of Tables | ix Acknowledgments | xi part 1 Thinking About Political Inequality | 1 1 Minority Groups and Political Equality in America | 3 2 Which Groups Govern? | 11 part 2 Differences in Representation | 37 3 Differences in Political Preferences and Priorities | 39 4 Differences in Policy Representation | 59 5 Differences in Legislative Representation | 77 part 3 Reducing Political Inequality | 117 6 7 8 9
Pluralism and Political Representation | 119 Descriptive Representation and Political Equality | 143 The Rewards of Voting | 173 Conclusion: The Future of Political (In)equality | 195
References | 209 Index | 221
Figures 2.1 White and African American conservatism and district racial composition | 29 2.2 Group differences in income, 1948–2003 | 34 3.1 Ideological orientation, African Americans, Latinos, and whites | 42 3.2 Bias in state ideology | 44 3.3 Predicted support for increasing spending in eight policy areas, whites and African Americans | 48 3.4 State-level bias on race-related issues | 50 3.5 Predicted support for increasing spending in eight policy areas, whites and Latinos | 52 3.6 Relative issue salience across groups | 55 3.7 Group salience of poverty, crime, and health insurance | 57 3.8 Impact scores for group salience | 58 4.1 Percentage of policy winners and “big losers,” whites and African Americans, 1972–2002 | 65 4.2 Percentage of policy winners and big losers, whites and Latinos, 1979–1994 | 69 4.3 Percentage of policy winners among high- and low-income respondents | 70 4.4 Percentage of policy winners and big losers, by income | 72 4.5 Percentage of policy winners and big losers, white and African American nonvoters | 74 5.1 Group differences in alienation | 78 5.2 Effect of group ideology on predicted senator w-nominate score, 107th Congress | 87 5.3 The possibility of perfect responsiveness without proximity | 89 5.4 Nondescriptive MCs and the proximity gap, 107th House | 93
viii | list of figures
5.5 Size of the proximity gap, various groups | 94 5.6 Nondescriptive MCs and the proximity gap in the 107th House, by district percentage African American or Latino | 96 5.7 The proximity gap if Congress is more conservative, liberal, or varied | 100 5.8 Nondescriptive MCs and the proximity gap on LCV votes, 107th Congress | 104 5.9 The proximity gap in the Senate, by state percentage African American or Latino | 104 5.10 The within-state proximity gap | 105 5.11 Party differences in the average distance of groups | 114 6.1 Percentage of policy winners in salient domains for minority groups | 123 6.2 Year-by-year analyses of African American and Latino policy winners on welfare spending | 125 6.3 Responsiveness of senators on issues distinctively salient for African Americans, 107th Congress | 132 6.4 Responsiveness of senators on issues uniquely salient for Latinos, 107th Congress | 134 6.5 The proximity gap on issues uniquely salient for minorities, 107th Congress | 137 6.6 Party differences in groups’ mean distance from MCs on salient issues | 139 7.1 Descriptive representation and the proximity gap | 148 7.2 The proximity gap and descriptive representation in districts at least 40 percent African American or Latino | 156 7.3 Descriptive representation and the proximity gap, salient issues | 159 7.4 Proximity to Democratic MCs on salient votes | 164 8.1 “Elections make the government pay attention to what the people think”: Difference in African Americans and whites responding “not much” | 180 8.2 Policy winners by group and turnout in three domains, 2000 | 183 8.3 Predicted ideological proximity of African American nonvoters and voters to MCs, by district percentage African American | 187 8.4 Rewards for voting in districts that are more than 40% African American or Latino | 188 8.5 Party differences in voting’s rewards | 192 8.6 Predicted distance of African American nonvoters and voters on LCCR votes, by district percentage African American | 193
Tables 5.1 Group differences in responsiveness, 107th Senate w-nominate Scores | 85 5.2 Models of distance from white MCs among co-partisan citizens and citizens represented by freshmen MCs | 99 5.3 Group differences in proximity, controlling for district heterogeneity | 102 5.4 State racial/ethnic composition and responsiveness, 107th Senate | 107 5.5 Individual-level models of responsiveness | 109 5.6 Party differences in relative responsiveness | 113 6.1 Votes used to create roll-call scores on salient issues, 107th Senate | 128 6.2 Models of LCCR votes, 107th Senate | 131 6.3 Models of NHLA votes, 107th Senate | 132 6.4 Votes used to create roll-call scores on salient issues, 107th House | 136 6.5 Party differences in responsiveness, salient issues | 138 7.1 Effect of descriptive representation, by party | 150 7.2 Descriptive representation and collective representation | 154 7.3 Descriptive representation and proportionality | 157 7.4 Effect of descriptive representation, salient issues | 162 7.5 Knowledge, interest, and descriptive representation | 168 8.1 Voting and ideological proximity, 107th Congress | 185 8.2 African Americans’ rewards for voting on salient issues, 107th Congress | 191
Acknowledgments Over the past four years as we have thought about political equality, we made many false starts, explored several dead-end alleys, and followed many a blind lead. To the extent that we finally got it right, we owe a great debt to the many people who corrected our path, pointed us in fruitful directions, and encouraged us to continue working on the important questions relating to political equality. Of course our thanking them does not implicate them in our errors, which are entirely ours. We are grateful to the careful readings, comments, conversations, and encouragement provided by Chris Achen, John Aldrich, Barry Burden, David Campbell, Dennis Chong, Martin Gilens, Paul Gronke, Zoltan Hajnal, Rodney Hero, David Lublin, Paula McClain, David Nickerson, Marvin Overby, Gary Segura, Jennifer Victor, Nicholas Winter, Christopher Wlezien, and Christina Wolbrecht. We also received helpful feedback from the participants of the Northwestern University American Politics Workshop, the University of Notre Dame American Politics Workshop, and the University of California–San Diego Legislative Studies Workshop. We are especially indebted to Larry Bartels, David Canon, Ben Page, and another anonymous reviewer for reading vast parts or all of the manuscript at one time or another. Their insights and encouragement were invaluable. As junior scholars, we are privileged to garner such attention and support from leading scholars. We also appreciate the efforts, encouragement, and assistance of John Tryneski and Rodney Powell of the University of Chicago Press. While working on this project, we received generous support from the University of Notre Dame Office of Research, the Institute for Scholarship in the Liberal Arts, for undergraduate research assistance and for summer funding; the University of Notre Dame Dean of Arts and Letters, for undergraduate research assistance; the offices of the Provost and the Dean of
xii | acknowledgments
Seaver College at Pepperdine University; and Duke University’s Program in Democracy, Institutions, and Political Economy. We are grateful to the several individuals and institutions for making their data publicly available. Without their efforts, we would have had much less to say. Thanks to Jeff Lewis and Keith Poole, the National Annenberg Election Survey, the Leadership Conference on Civil Rights, the League of Conservation Voters, and the National Hispanic Leadership Agenda. Thanks also to Mallory Brown and Darlene Luebbert for expert research assistance. Portions of our analyses were published in “The Unequal Representation of Latinos and Whites,” Journal of Politics 69(4): 1032–46 (2007). Brian thanks his whole family for unending support and prayers. To Dina, thanks for everything, from giving me all the time in the world to hatch this project in Elma to the bigger challenge of carving out time for me to carry it through in Malibu. Thanks for listening, reading, watching, walking, and thinking with me. To Caleb and Alexander, born in the midst of this project, all my love—more each day. John thanks his parents for encouraging him to think about inequality from an early age. To Natalie, Jack, and Elizabeth, thanks for the bike rides to school, hugs at the door, and playing with me in the yard. And to Amy, thank you for living through the many ups and downs of this long project, and for providing the advice, nudging, and support to make it happen.
p a r t one
Thinking About Political Inequality
one
Minority Groups and Political Equality in America In late summer 2001, four American passenger planes were hijacked by terrorists, resulting in the destruction of the World Trade Center in New York City, extensive damage to the Pentagon building in Washington, D.C., and a plane crash in western Pennsylvania. All told, 3,043 people died as a result of these actions. Among those killed in New York, 76% were white, 9% were Latino, and 8% were African American.1 Eleven days after these tragic events, Congress created the September 11th Victim Compensation Fund. In total, the fund distributed more than $7 billion to survivors of 2,880 persons killed in the attacks and 2,680 persons injured by them (Feinberg 2004, 1). The average award to families of victims killed in the attacks was greater than $2 million. The average award to injured victims was more than $400,000.2 In late summer 2005, Hurricane Katrina, a category 4 tropical storm, struck the coast of the Gulf of Mexico. Its storm surge breached the levee system that protected New Orleans from Lake Pontchartrain and the Mississippi River, and most of the city was subsequently flooded. This and other major damage to the coastal regions of Louisiana, Mississippi, and Alabama made Katrina the most destructive natural disaster in the history
1. Schwartz et al. 2002. See table 1 at the Web site of the Centers for Disease Control Weekly Morbidity and Mortality Weekly Report, at http://www.cdc.gov/mmwr/ preview/mmwrhtml/ mm51SPa6.htm#tab1. These data include foreign-born victims, so the domestic racial composition of the September 11th victims likely was even more skewed toward whites. 2. Private contributions boosted the average compensation level even higher. According to a report by the RAND Corporation, families of civilians received an average of $3.1 million in government and charitable awards, while families of those who died in uniform received an average compensation award in excess of $4.2 million (Dixon and Stern 2004).
4 | chapter one
of the United States, with property damage totaling approximately $125 billion.3 The storm’s official death toll exceeded 1,300, with one estimate from the Louisiana Department of Health showing that 49% of the state’s deceased storm victims were African American, and 45% were white.4 More than a million people were displaced—a humanitarian crisis on a scale unseen in the United States since the Great Depression. As of November 22, 2005, the federal government estimated that it had provided more than $4.4 billion to 1.4 million families affected by the hurricane, or about $3,100 per family.5 These were rather different kinds of disasters, but many drew important comparisons between the two. Several African American leaders, for example, criticized what they saw as disparate federal responses to these tragedies. Jesse Jackson, president and founder of the RainbowPUSH Coalition; Marc Morial, president of the National Urban League; Bruce Gordon, former president of the National Association for the Advancement of Colored People (NAACP); and the Congressional Black Caucus all called for the creation of a government fund similar to that which compensated September 11th victims. In a press release, Gordon argued, “You got people who lost their jobs, homes and had all of their assets depleted. In some cases, families lost their bread winner. If the victims of 911 deserved compensation, and they did, then certainly these victims deserve no less to help restart their lives.”6 Morial said in a press release, “As it did then, Congress must take immediate and decisive action to begin compensating American citizens whose lives have been disrupted by this major national tragedy.”7 These spokesmen for the African American community were asking, in the words of the September 11th Compensation Fund Special Master Kenneth Feinberg, “Why should Congress provide very generous compensation to a very limited number of individuals while excluding other victims of life’s misfortunes?” (Feinberg 2004, 78).
3. See “Money” at the USA Today Web site: http://www.usatoday.com/money/economy /2005-09-09-katrina-damage_x.htm. 4. See the Web site of the Louisiana Dept. of Health and Hospitals at http://www.dhh .louisiana.gov/offices/news.asp?ID=145&FromSearch=1&Detail=746. 5. See the news release on the Federal Emergency Management Agency (FEMA) Web site: http://www.fema.gov/news/newsrelease.fema?id=20818. 6. See the NAACP Virginia State Conference Web site: http://www.virginianaacp.org/ news/cco/victimsfund.htm. 7. See the page for 2005 press releases at the Web site of the National Urban League: http://www.nul.org/PressReleases/2005/2005PR222.html.
minority groups and political equality in america | 5
our project and its significance This is not a book about September 11th or Hurricane Katrina. However, in the aftermath of the disparate federal response to the twenty-first century’s two most significant disasters in the United States, many are asking questions about the federal government’s even-handedness in constructing public policies that affect different racial groups. Is the response to Katrina merely the tip of the iceberg? Is it an indication of widespread inequality in Washington’s day-to-day operations? We do not have to look very hard to identify other recent policies that reflect apparent inequality in racial groups’ political clout. In 2005 alone, African American opposition was unable to block passage of new voter identification requirements in Georgia; an overhaul of the nation’s bankruptcy system that leaves the underprivileged exposed to predatory lending; cuts in entitlement spending, including Medicare and Medicaid in the fiscal 2005 budget—and the list goes on. Whether these policies are good or bad is a matter for debate. What is clear is that African American opposition to these policies did not stop them from being enacted. In this book we ask, “Which groups govern?” That is, which groups get what they want from government? How much more does government action respond to the demands of the majority white population than it does to minority group demands? This is a critical question to ask in the United States, a racially and ethnically diverse nation that has been challenged by deep racial divisions since before its inception. We will also investigate various avenues to equalize groups’ representation. Of course, American society is not simply made up of African Americans and whites. The U.S. Census 2005 American Community Survey estimates that Latinos comprise 14.5% of the nation’s population, a proportion that is growing quickly. To understand politics in the United States today, we analyze whites, African Americans, and Latinos, now America’s largest minority group. Since studies of minority politics are only beginning to examine more than one or two groups at once (e.g., McClain and Karnig 1990; Kaufmann 2003; Meier et al. 2004), we have much to learn about the ways multiple racial and ethnic groups in the United States relate politically. We assess the extent to which government action and policy reflect these three groups’ preferences. In particular, we gauge these groups’ representation relative to each other, asking how much more whites are represented in political outcomes and examining the conditions under which minorities and whites are more equally represented. In analyzing
6 | chapter one
the relative representation of these groups, we look at various points in the American system, including roll-call votes in the House of Representatives and the Senate, and the content of federal public policies forged across the system. We begin by exploring exactly what political equality means. Defining equality turns out to be no easy task. We identify three different standards of political equality that have influenced political theory and governmental practice in the United States. First, equality may mean that each group’s likelihood of being on the winning side of political decisions should be proportional to its size. This proportionality standard is clearly illustrated by the principle of “one person, one vote.” On the other hand, equality may require that minority groups should be better represented in outcomes than their numbers would suggest. We call this the race-conscious egalitarian standard. Although it is the most stringent standard and is often controversial, the federal government has adopted it from time to time, and some prominent legal and political minds continue to argue for its application. Finally, equality may mean that various groups in society are better represented in areas of unique importance to them. This is the pluralist standard, one that has deep historical roots in the United States. As we might expect, our assessment of the degree of equality changes as we alter the standards of evaluation. Taking these three standards as our benchmarks, we examine the extent to which whites and minority groups are equally represented in American politics. Scholars have been concerned with the representation of minority groups for decades. Many have examined the factors that might heighten or depress the political clout of minority groups, including the racial and ethnic composition of electoral districts, the race and ethnicity of elected officials, and the degree of minority participation allowed or mobilized in electoral contests. Rarely, however, have studies explicitly examined the representation of various groups in relation to each other. In the pages that follow, we report direct measures of the extent to which white citizens are better represented in political outcomes than are Latinos and African Americans, as well as how various factors affect these relationships. From one perspective, our research documents considerable inequality of representation in American politics. Our analyses show that congressional votes and the ultimate content of public policy generally are much more in line with white Americans’ preferences. In fact, in many cases minority groups are not better represented even where they make up a larger proportion of electoral districts, suggesting that minorities as individuals are unequally represented.
minority groups and political equality in america | 7
However, there is much more to the story. Three factors can boost minorities’ political representation, sometimes enough to rival that of whites. First, African Americans and Latinos make gains relative to whites and are sometimes even better represented than whites on issues that these groups care about more than whites do. Second, when represented by members of their group (i.e., when minority constituents are descriptively represented), Latinos and African Americans are represented just as well as whites, at least under some conditions. Third, getting out the vote can provide a small boost to minorities’ relative political representation in certain contexts. However, voting by itself does not enhance minorities’ relative representation enough to produce equality. These three findings hold important implications for strategies to achieve greater equality of representation among these three groups. In short, as we shall show, the story of political equality in America is nuanced, multifaceted, and depends in many ways on one’s chosen definition of political equality. We believe that improving our understanding of the extent and causes of disparities in political representation is essential. For, while Americans widely value the abstract principle of equality (e.g., McClosky 1964; Hochschild 1981), when it comes to the details, “disparities in political influence, even if deplored, [are] widely tolerated” in American politics (Verba and Orren 1985, 20). Yet there are a number of reasons to be concerned about the equality of minority groups’ political representation. First, equal representation is central to democracy (Dahl 1961). According to a leading scholar of political participation, “the equal consideration of the preferences and interests of all citizens” is “one of the bedrock principles in a democracy” (Verba 2003, 663). Stated less stringently, “[d]emocracy implies a certain degree of political equality—if not full equality of political representation among citizens, at least some limit to political inequality” (Verba and Orren 1985, 8). As the Katrina experience shows, another reason that we might strive for equal political representation among citizens is that political representation can protect citizens’ other basic rights. Political equality is tied to other forms of equality, such as economic and social equality. If minority groups are not politically equal, it is more difficult for them to protect their economic and other interests. Given these stakes, it is not surprising that minorities care deeply about equality in various forms. For example, in the NAACP’s mission statement, the first “primary objective” listed is to ensure “the political, educational, social and economic equality of all citizens.”8 8. See the NAACP Web site at http://www.naacp.org/about/mission.
8 | chapter one
Finally, measuring and explaining the political equality of minority groups is important if we are to fashion public policies to enhance political equality. That is, we must know how much political inequality exists and its sources in order to formulate strategies to promote greater equality. There is some disagreement over appropriate terms to use when referring to African Americans, Latinos, and whites. Throughout the text, we generally refer to African Americans as a “racial” group and Latinos as an “ethnic” group, though we realize these are imperfect categories. We try to be sensitive to the diversity within all the groups, reporting, for example, differences among Latinos of different national origins. In general, we have attempted to strike a balance between precision and avoiding excessive wordiness, sometimes using “minority groups” or “minorities” to refer to both African Americans and Latinos. We also note that we treat these categories as mutually exclusive. We provide further details on the specific operationalization of these groups in later chapters.
what lies ahead The next chapter lays the conceptual and theoretical groundwork for the rest of the book and summarizes what we do and do not know about political equality in America. Specifically, we define what we mean by political representation, introduce the sphere of political equality in which we operate, and elaborate on the three standards of equality introduced above. We also explain from a theoretical standpoint why minority groups might be unequally represented and why this may change as the racial and ethnic composition of districts varies, as we look at different issues, and as we consider factors like voter turnout. Part 2 (chapters 3–5) describes differences in the political representation of African Americans, Latinos, and whites evaluated against the standards of political equality we have articulated. Chapter 3 lays a foundation for this by showing that these groups often have different policy preferences and priorities, and that these different attitudes are not reducible to income differences among the groups. Different preferences create the possibility that some groups are better represented than others. Different priorities create the possibility that these groups’ representation depends on the relative salience of issues across the groups. Chapter 4 asks how much more often whites obtain their preferred policy outcomes than do African Americans or Latinos. It shows that when the groups disagree about the direction of federal policy, whites are
minority groups and political equality in america | 9
more likely to see their preferences implemented on most issues. These analyses show that minorities are not as well represented as whites and therefore are politically unequal under a strict, race-conscious egalitarian standard. Chapter 5 probes whether differences in the representation of the groups’ opinions in policy are tied to differential representation in legislators’ rollcall votes. It shows that whites are better represented than minorities in legislators’ roll-call votes in most issue areas. In this chapter we also investigate the degree to which the proportionality standard is met by examining whether minority groups are equally represented when they comprise a substantial share of an electoral district’s population or nearly half of the population. We demonstrate that larger African American populations in electoral districts do not improve African Americans’ relative representation, which violates the proportionality standard. However, we find some evidence that Latinos are equally represented under the proportionality standard. Part 3 (chapters 6–8) examines three ways to narrow the gaps in political representation. In chapter 6, we move to the third standard of political equality, pluralism. In contrast to chapters 4 and 5, where we focus on issues that are equally salient to the three groups or are less salient to the minority groups than to whites, in chapter 6 we look at issues that are more salient to minority groups than they are to whites. In these policy domains, we find that minority groups’ representation often rivals that of whites, both in terms of legislators’ decisions and government policies. Therefore, the pluralist standard is generally satisfied. Chapter 7 examines a second way to close the representation gap: electing minority representatives, thereby establishing descriptive representation. This chapter shows that descriptive representation sometimes levels the disparities in representation, yielding political equality. However, although descriptive representation brings representation gains in some contexts, in others it offers few gains beyond the benefits minorities enjoy from being represented by Democrats. Finally, chapter 8 asks how political participation affects the groups’ political representation. In general, political activity increases one’s political representation (Griffin and Newman 2005). However, we find evidence that African Americans and Latinos generally garner fewer rewards for voting than do whites. There are some conditions under which the rewards of voting are equal among the groups, but these tend to be the exceptions rather than the rule.
10 | chapter one
We close in chapter 9 by pulling together the various findings. Our analyses point to some tentative predictions about the future of political inequality in American politics. They also hold important implications for strategies and policies designed to enhance the political equality of minority groups.
tw o
Which Groups Govern? There is nothing more dangerous than to build a society, with a large segment of people in that society, who feel that they have no stake in it. dr. martin luther king jr., 1965 commencement address at Antioch College
For a group to have a stake in a political system, it must believe that it has some clout in the system’s decisions. Otherwise, little would be lost if that manner of organizing the society’s decisions were replaced with another. Do African Americans and Latinos have a stake in the U.S. political system, in the sense that their interests are converted into governmental action? Is this stake equal to that of whites? This chapter defines some of our key terminology and lays the theoretical groundwork for our subsequent empirical investigations. We spend much of the chapter delineating the concepts of representation and political equality. We introduce various spheres of political equality and identify the type of political equality in which we are most interested—the equality of intermediate and final political outcomes. Thereafter, we describe three standards of political equality by which to judge whether African Americans, Latinos, and whites are appropriately represented in these outcomes. This allows us to place our investigation in the context of prior work and to develop theoretical expectations about relative representation among these groups. We close the chapter by pointing out the unique contribution our study makes by focusing on the relative equality of these three groups.
12 | chapter two
spheres of representation and political equality Representation, a fundamental element of democracy, is a rich and multifaceted concept. Scholars have penned entire books outlining the various aspects of this concept (e.g., Pitkin 1967; Manin 1997). We focus on one dimension of representation, the confluence between what actors in the government do and what citizens want them to do. If a citizen’s preferred course of action for the government is consistent with what the government subsequently does, we conclude that the citizen is well represented, that is, better represented than a citizen whose preferences are not reflected in governmental action. Our focus is on describing how well citizens’ views on policy are represented in the outcomes of the political process: the actions of elected officials and the policies they enact. We assume, safely we think, that any group of citizens would prefer to see policy outcomes reflect its views. We also note that to be represented in this sense does not require either that a group of citizens affirmatively attempt to change the direction of public policy or that elected officials be aware of and give special weight to the preferences of a group for the group to be represented. It only requires that the group’s views prevail when the government makes decisions. For instance, if citizens tend to vote for likeminded representatives, and officials merely consult their own preferences when they vote, officials’ decisions will tend to reflect voting citizens’ preferences. We consider this representation, even though these citizens may not have directly caused their representatives to reflect their views by bringing pressure to bear on the representatives when it comes time to cast a roll-call vote. Although we may want to know why some citizens are better represented than others (indeed, we examine this question in later chapters), from a citizen’s perspective, so long as the government or its actors are making decisions with which the citizen agrees, the citizen benefits. Equality of representation in political outcomes is just one important form of political equality. Most generally, “political equality refers to the extent to which citizens have an equal voice in governmental decisions” (Verba 2003). More precisely, political equality demands first of all that individuals possess equal legal rights to participate in the “inputs” of the policy process: elections and other means of communicating their preferences to government officials (Verba, Schlozman, and Brady 1995; Verba 2003).1 1. A more demanding requirement is that all citizens have equal ability to participate in conveying their preferences to elected officials, requiring much greater parity in citizens’ resources (Verba 2003).
which groups govern? | 13
America has a long history of denying this type of political equality to minority groups. African Americans were not guaranteed the franchise until the ratification of the Fifteenth Amendment, and the Jim Crow laws severely limited voting by African Americans in the South until the 1965 Voting Rights Act (VRA) (Kousser 1974 and 1999). Many Latinos in the American southwest also were effectively disenfranchised until the passage of the Voting Rights Act . The struggle for this type of political equality is ongoing. For example, felon disfranchisement laws in some states disproportionately prohibit African Americans from voting; a recent Georgia law requiring photo or digital identification to vote was overturned as racially discriminatory; efforts to challenge voters’ eligibility often appear to be targeted at African American and Latino localities; and renewal of the VRA, parts of which were set to expire in 2007, sparked some controversy (e.g., Lyman 2006).2 Political equality also demands the equal consideration of citizens’ inputs—none should be “weighted” more than others by the rules of politics.3 For instance, the U.S. Senate’s apportionment scheme and, by extension, the electoral college both give greater “voting weight” to the citizens of less populated states, which means that politically relevant groups concentrated in more populated states, including African Americans and Latinos, are disadvantaged in these bodies (Dahl 1956; Lee and Oppenheimer 1999; Griffin 2006). Institutional features of the political system like these that give greater per capita influence to some citizens on arbitrary grounds also create political inequalities among citizens ( Verba 2003). A final requirement of political equality relates more directly to citizens’ relationship to governmental “outputs.” Namely, political equality demands “equality in [governmental] response” ( Verba 2003, 665). In this book, we focus on this final type of political equality among groups—whether they are equally represented in the decisions made by elected officials and the government as a whole. Of course, on any single policy decision, “equal treatment for all is not possible, since individuals and groups have different needs and preferences, and policies favoring some are less favorable to others” (666). However, across multiple policy decisions we can weigh 2. One commentator went so far as to argue that African Americans should be sure to vote in the 2004 elections, saying, “If you don’t vote on Nov. 2, you may not get another chance.” See “8 Reasons Why Black People Should Vote,” at http://www.blackamerica web.com/site.aspx/bawnews/impact04/reasons. 3. This is reflected in the “one person, one vote” principle. See Wesberry v. Sanders, 376 U.S. 1 (1964).
14 | chapter two
whether some groups are on the winning side more often than others, or whether some are better represented in terms of outputs.
standards of political equality While there is general agreement that citizens should have equal opportunities to participate in politics and that these inputs should be weighted equally, what it means for African Americans, Latinos, and whites to be equally represented in government action is contentious. As Verba, Schlozman, and Brady (1995, 12) note, “If political equality is hard to achieve, it is almost as hard to define.” How often should these groups get what they want before we can conclude that they are politically equal? There is no obvious answer to this question. At least there is no consensus on which answer is best. Various theorists have offered different answers. Therefore, even though many may believe that minority groups are unequally represented in the United States, these impressions assume a standard of equality that others may not share. As we will see, our conclusions about the extent of political inequality change depending on what is meant by political equality. What may seem to be the simplest formulation of equality, that each group should be equally represented, is actually deeply problematic because minority groups by definition are numerically smaller than the majority white group. If groups are to be strictly equal, this would require representatives in states or congressional districts where minorities make up only 5% of the population to represent minorities and whites equally well. For example, in North Dakota, where African Americans, Latinos, and whites make up 0.8%, 1.4%, and 92% of the population, respectively, strict equality across groups would require North Dakota’s Senators to represent these groups equally.4 But few would argue that groups consisting of less than 2% and 92% of the constituency ought to be equally represented. Moreover, few would expect reelection-minded representatives to act in such a way. What then might it mean for groups to be politically equal in terms of outputs? Various political theorists, legal scholars, and politicians have argued for three different standards to evaluate equality of representation—standards that we and they have termed proportionality, race-conscious egalitarianism, and pluralism. A first standard by which we might judge the 4. Percentages are based on the 2005 American Community Survey, North Dakota Fact Sheet, U.S. Census (Washington, DC: U.S. Govt. Printing Office).
which groups govern? | 15
political equality of government decision making is that of proportionality. “The basic characteristic” of proportionality is that “all groups influence a decision in proportion to their numerical strength” (Steiner 1971, 63; Mansbridge 1999, 634). By extension, proportionality requires that each individual should be equally represented regardless of the racial, ethnic, or other group to which that person belongs. As Sidney Verba described this standard, “Political equality refers to the extent to which [individual] citizens have an equal voice in governmental decisions” (2003, 663; see also Achen 1978, 497). We see the “input” side of this notion most clearly enshrined in the principle of “one person, one vote” established by the Supreme Court in Wesberry v. Sanders. Although that case dealt specifically with the size and shape of legislative districts, the general principle of the equality of individuals at its core is a central principle in American political thought.5 Our study pertains to the less-often studied equality of government outputs, or whether outcomes are also proportional. In this view, evidence that whites’ preferences are generally better reflected in government action than are the preferences of minority groups may suggest, but does not prove, that the groups are politically unequal. Instead, the proportionality standard requires evidence that individual African Americans or Latinos are less well represented than individual whites, or that whites’ representation as a group is disproportionate to its group size.6 As applied to the relative representation of minority groups, the proportionality standard requires that where the African American or Latino share of an electoral district’s population is large, representation of African Americans or Latinos should move toward equality with whites. Ultimately, according
5. Putting this another way, one legal scholar argued that “when [majoritarian] institutions . . . produce outcomes that result in differential provision of public goods and services to political minorities (for example, in the rural South fewer roads being paved in the black side of town)—it seems difficult to resist the conclusion that Madison’s nightmare of ‘majority factionalism’ has become a reality” (Pildes 2000, 119–20). In other words, a proportional view argues that the ratio of paved roads to unpaved roads ought to be the same in “the black side of town” as in “the white side of town.” 6. In Lani Guinier’s view, “each voter should enjoy the same opportunity to influence political outcomes. No one is entitled to absolutely equal influence; by the same token, no one is entitled to grossly disproportionate influence or a monopoly on control. The majority should enjoy a majority of the power, but the minority should also enjoy some power too. Thus, proportionate interest representation measures opportunities for fair participation using a baseline of proportional power or proportional influence” (1994, 25).
16 | chapter two
to the proportionality standard, in an electoral district comprised of two groups of the same size, the two groups should be equally represented. This standard for judging the equality of political outputs is likely to be relatively unobjectionable. First, a standard of equality that judges outputs based on how they affect individuals as opposed to groups should be appealing given the strong sense of individualism among Americans ( Verba and Orren 1985, 6; Feldman and Zaller 1992). Second, most Americans value political equality, especially in terms of inputs, but to some extent outputs as well. “In [American] politics the ideal is democracy and one person, one vote. This ideal condemns gross disparities of political power among individuals and groups, and hence approaches equality of result” (Verba and Orren 1985, 8; emphasis added). Third, even those who advocate a more stringent standard of equality would presumably agree that political equality demands at least proportionality. How can we know if citizens are equal in a proportional sense? We can examine whether minority groups’ representation improves where those groups comprise greater proportions of the constituency. In a related way, we can focus on locations where two groups of roughly equal size disagree about the direction of public policy to see which group wins. We adopt both of these approaches below. For some commentators, the proportionality standard does not go far enough; they argue for a more stringent standard that we refer to as raceconscious egalitarianism (Fiss 1976; Colker 1986; MacKinnon 1987; Tribe 1988; Sunstein 1994; Harris 2000; Issacharoff and Karlan 2003; Smith 2003). This view, sometimes associated with the “antisubordination principle,” argues that contemporary American society includes some “specially disadvantaged groups,” including African Americans and Latinos, among others (e.g., Fiss 1976, 155). These groups have long occupied “the lowest rung” of the social ladder and “they are America’s perpetual underclass” according to Yale law professor Owen Fiss, a leading proponent of this view (1976, 150). These writers contend that laws should work to protect and advance the status of these disadvantaged groups. For example, Fiss has argued that government policies should aim not simply for the equal treatment of individual citizens, but for the facilitation of equal status among groups (176). To this end, Fiss and others contend, laws that harm disadvantaged groups, even unintentionally, violate the Fourteenth Amendment’s Equal Protection Clause and should be deemed unconstitutional. Fiss contends that “any state action that systematically creates, aggravates, or perpetuates the subjugation of blacks . . . would constitute a violation of equal protection” (2004, 10).
which groups govern? | 17
This view explicitly conceives of equality in group terms, rejecting what Fiss calls “a highly individualized conception of rights” (1976, 127), a conception at the heart of the proportionality standard. For example, Rogers Smith (2003) implicitly rejects Justice Clarence Thomas’s claim that in the Equal Protection Clause “lies the principle that the government must treat citizens as individuals, and not as members of racial, ethnic, or religious groups” (1).7 Smith contends “it seems most appropriate to focus on whether government is succeeding in providing basic substantive protections to all groups as well as individuals in American society”(1–2).8 To advocates of race-conscious egalitarianism such as Smith and Iris Marion Young (2002), equal protection of individuals is not enough simply because the average individual who happens to be African American or Latino has not enjoyed the economic, social, and political status of the average white individual and currently is still in a disadvantaged status. Proportional decision making is thus not enough to guarantee political equality because proportionality keeps minority views subject to the majority’s will. To this end, status-disadvantaged groups often should receive benefits from government decisions that are out of proportion to their size. All the more so, argue Smith and Young, because the American political and legal systems have in large measure created these groups’ disadvantaged status. From this perspective, since whites have perpetuated racial categories to their group’s advantage, creating an underclass that is a numerical minority, it is unfair for whites today to fall back on majority rule at every turn as the method of aggregating preferences. Therefore, race-conscious egalitarianism demands more than merely proportional representation. In this view groups (perhaps with the qualifications that they be of some minimum size and have experienced historical discrimination) should be “more than proportionally” likely to see their preferences realized in the decisions of government. If African Americans, Latinos, and whites are to be political equals, according to this stance, the preferences of the minority groups must bear special weight. So, if a racial or ethnic minority group comprises 10% of a district, advocates of this view might argue that the group should be treated as if it comprised 20% or 30% of the district. As the size of the group grows, producing more than
7. Missouri v. Jenkins, 515 US 70, 120-21 (1995) (J. Thomas, concurring). 8. As Justice Lewis Powell recognized, “The concept of ‘representation’ necessarily applies to groups: groups of voters elect representatives, individual voters do not” (Davis v. Bandemer, 478 U.S. 109, 167 [1986]).
18 | chapter two
proportional outcomes leads ineluctably to outcomes in which groups are treated equally. In sum, minorities might merit equal treatment when they comprise some substantial proportion of an electoral district, even if this proportion is smaller than the proportion of whites. The race-conscious egalitarian view is not simply an abstract principle. Prior to 1976, several federal court decisions found that government policies creating racially disparate impacts violated the Fourteenth Amendment’s equal protection requirement.9 That is, government policies had to provide equal benefits or costs to various racial and ethnic groups. Although the Supreme Court has altered this requirement as a constitutional matter,10 the Court has indicated that Congress may pass laws banning policies that disproportionately harm minority racial and ethnic groups, and Congress has proceeded to do so. For instance, Title VII of the 1964 Civil Rights Act now allows employment discrimination to be established by proof of discriminatory impact, and the 1982 Amendments to the VRA permit proof of discriminatory impact to establish a violation of that law.11 So, for example, if an employer uses a hiring criterion that on its face is race-neutral (e.g., a standardized test) but results in no minority employees or even candidates, that employer may be found to have violated Title VII. Or a state law that required proof of identification to vote could be challenged under the VRA on grounds that the implementation of this facially neutral law disproportionately harms African Americans. Furthermore, arguments for the adoption of a race-conscious egalitarian standard of equality have not been confined to a legal setting. When legal scholars argue that laws disadvantaging minorities should be deemed unconstitutional on equal protection grounds, they are not only backing a change in behavior by the courts, they are implicitly advocating a change 9. See, for example, Primus (2003); Castro v. Beecher, 459 F.2d 725, 732–33 (1st. Cir. 1972); Chance v. Bd. Of Exam’rs, 458 F.2d 1167, 1176–77 (2d. Cir. 1972). 10. In Washington v. Davis, 426 U.S. 229 (1976), the Supreme Court ruled that absent evidence of discriminatory intent, facially neutral governmental actions that disadvantage racial minorities are not subject to special judicial scrutiny to determine their constitutionality. However, prominent legal scholars continue to argue for this standard. For example, Tribe (1988, 1516–19) argues that “[m]inorities can also be injured when the government is ‘only’ indifferent to their suffering or ‘merely’ blind to how prior official discrimination contributed to it and how current official acts will perpetuate it.” Issacharoff and Karlan claim that in Washington v. Davis the Court exhibited a “cramped notion of equality” (2003, 3). 11. See Griggs v. Duke Power Co., 401 U.S. 424 (1971) and Thornburg v. Gingles, 478 U.S. 30 (1986).
which groups govern? | 19
in legislative behavior. That is, if legislatures ceased passing laws exacerbating the subordination of disadvantaged groups, courts would not need to rule these laws unconstitutional. Indeed, scholars have explicitly contended that the antisubordination principle (sometimes called the anticaste or group-disadvantaging principle) speaks to legislative outcomes. For instance, in his seminal article on antisubordination, Fiss (1976, 153) argued that “the judiciary is attempting to rectify the injustice of the political process as a method of adjudicating competing claims.” Iris Marion Young (2002, 9) similarly stressed that “courts cannot be the main arenas in which to combat continuing status inequality. . . . [Fiss] calls upon us to notice and theorize status inequality, argue against it to our fellow citizens, and try to change it. Legislative action may be at least as useful as litigation in doing that.” Issacharoff and Karlan (2003, 17) take this even further, arguing that the Court’s rejection of the antisubordination principle has left it to the legislature to pick up the slack: “[O]nly civic and legislative commitment to the group-disadvantaging principle’s capacious understanding of equality has kept it from being eclipsed altogether by the Supreme Court’s narrow focus on individual interests.” Finally, in defining his “anticaste” principle, Cass Sunstein (1994, 2436, 2440) points out that “[w]e can understand this principle as emphasizing legislative rather than judicial duties,” and further that “[i]f the legal culture is to return to the roots of the constitutional commitment and to a better understanding of equality, the legislative branch should take the lead.” What might the race-conscious egalitarian standard look like in practice? On any single vote, the only way for two groups to be equally represented would be for groups to share the same preferences or for one of the roll-call alternatives to lie directly “between” the positions two groups have staked out. However, across multiple votes, elected officials can respond on some votes to the preferences of one group and on other votes to the preferences of another group. Doing so in a way that reflects minorities’ preferences substantially more often than is proportional to their group size would satisfy this standard of equality. The race-conscious egalitarian view surely is the most stringent standard of equality, one with which many may disagree or desire to restrict.12
12. Consider the extreme form of this argument, which would say that five hundred Native Americans or twenty Laotians in a given congressional district should have substantial say in the formation of policy. Some have argued that application of the egalitarian standard is more compelling with respect to groups that have been historically
20 | chapter two
For example, even Lani Guinier (1994, 14), the Clinton nominee to head the Justice Department’s Civil Rights Division who was labeled by her opponents as the “quota queen,” rejects strict race-conscious egalitarianism in favor of something akin to proportional outcomes, arguing that “the purpose [of Guinier’s proposed electoral reforms] is not to guarantee ‘equal legislative outcomes’; equal opportunity to influence legislative outcomes regardless of race is more like it.” Race-conscious egalitarianism is not the present constitutional law of the land, and some may deem that it gives undue influence to minority groups. Nevertheless, it is a standard of equality that has been advocated by noted scholars and applied by the courts and by legislatures in a number of situations involving racial and ethnic groups. In the context of the relative representation of African Americans, Latinos, and whites, we will test whether American politics meets the yardstick of race-conscious egalitarianism by assessing whether, as groups, African Americans and Latinos find that their preferences are more than proportionally represented in the decisions of government compared to those of whites, even when whites outnumber African Americans or Latinos in an electoral district. We sometimes focus on whether groups are equally represented because it is a clear benchmark against which to judge the relative representation of these groups. Thus, if minority groups are not equally represented, they are unequal under a very strict race-conscious egalitarian standard of political equality. When feasible, we then loosen this standard to a “more than proportional” comparison, examining whether minority groups are represented as well as whites when the minority groups constitute at least 25% of a state or district population. If so, we conclude that these groups are more than proportionally represented. A third standard of political equality, pluralism, argues that intensely held preferences ought to be specially represented, even when those preferences are in the minority.13 The “fundamental axiom” of pluralism is that “[i]nstead of a single center of sovereign power there must be multiple disadvantaged by state-sanctioned discrimination (Anaya 1997). For, “while ethnic or cultural groups may not be intrinsically superior to other types of groups (such as associations of skiers or Porsche drivers), they may in some sense be more important” (222). We could add to this the requirement that a group must constitute a minimal proportion of an electoral district to merit equal representation. 13. The opposite view, which Kendall and Carey (1968, 10) term “populistic democracy,” contends that if a minority has intense preferences, “[s]o much the worse for the minority. If it feels so strongly about the matter, let it get out and win a majority over to its side.”
which groups govern? | 21
centers of power, none of which is or can be wholly sovereign. Although the only legitimate sovereign is the people, in the perspective of American pluralism even the people ought never to be an absolute sovereign; consequently no part of the people, such as a majority, ought to be absolutely sovereign” (Dahl 1967, 24). From the pluralist perspective, the political system is divided into numerous policy domains and “who governs” varies in each domain (Polsby 1980). Pluralists therefore generally feel that “the political system [is] open to multiple interests if these interests feel strongly enough about an issue to mobilize pressure” (Manley 1983, 369). That is, intense minorities can exert significant influence in issue domains of unique importance to them. Thus, pluralism is in part an answer to what Dahl called “the intensity problem,” which arises when a majority that does not care about an issue opposes a minority that cares deeply about it. Dahl (1956, 90) dramatically claimed that “Madison might argue that government should be designed to inhibit a relatively apathetic majority from cramming its policy down the throats of a relatively intense minority.” One reason that we might want to solve the intensity problem is that, according to some, no democracy is viable that fails to take it into account (Kendall and Carey 1968, 10). Another reason is fundamental fairness. According to Mayo (1960, 178), “to count each person . . . equally is absurd: some people feel more strongly about certain issues than others. Would it not fly in the face of common sense and elemental fair play to argue that 50% plus one of the lukewarm should overrule 50% minus one consisting of passionate dissenters?” Oddly enough, an early proponent of a form of pluralism sought to use this principle to suppress the rights of racial minorities. In his A Disquisition on Government, once Congressman, Secretary of War and State, and Vice President John C. Calhoun argued for an interpretation of the constitution that included “concurrent majorities.” Writing before the Civil War, Calhoun proposed that a majority from any geographic region (including especially the American South) should have the constitutional power to veto acts of the federal government. Specifically, to avoid a “majority faction,” one should “tak[e] a sense of each interest or portion of the community which may be unequally and injuriously affected by the action of the government . . . and . . . require the consent of each interest either to put or to keep the government into action” (Calhoun 1953, 20; emphasis added). Only by using this mechanism will “every interest . . . be truly and fully represented” (21). Pluralism has been a “dominant theory or paradigm of power among American social scientists” (Manley 1983, 368). However, its normative
22 | chapter two
status is not unquestioned. Some fear that pluralism may be inadequate to protect disadvantaged minority groups. For instance, Schattschneider (1960, 34–35) famously claimed that “the flaw in the pluralist heaven is that the heavenly chorus sings with a strong upper-class accent,” meaning that the socially advantaged have much greater influence in the actual workings of government. More generally, “not all groups in the pluralist United States are equal” (Manley 1983, 376). This less sanguine view of pluralism would anticipate that perhaps even on the issues they care more about than whites, minorities may still be unequally represented. Critics of pluralism might also protest that it does not go far enough, arguing that minority groups ought to be more than proportionally represented on a broad range of issues, not just on their “pet issues.” Given pluralism’s prominent place in political thought, we pay attention to differences in relative representation across issue domains. In our analyses of minority groups’ representation, if minorities are more than proportionally represented on issues that minorities care more about than whites do, we conclude that this standard of political equality is satisfied. Stated another way, if the standard of race-conscious egalitarianism is met in domains that are more important to African Americans or Latinos than to whites, we take this as evidence that the pluralist standard of equality is met. In the end, we leave for normative scholars the task of sorting out how well numerical minorities, especially racial and ethnic minorities, should be represented. Our purpose here is not to advocate one of these standards over another. We simply aim to determine in subsequent chapters whether any of these standards of equality is met with respect to African Americans, Latinos, and whites in the making of public policies, and how various conditions affect these groups’ equality relative to these standards.
who governs? Our focus on groups’ relative representation naturally leads us to join a long line of investigations into a core question of politics: Which groups in society enjoy the greatest say in the government’s decisions (e.g., Dahl 1961; Polsby 1980; Lowi 1969; Lindblom 1977)? A series of studies have shown that citizens as a whole are effectively represented—that when citizens’ support for a policy changes, government tends to respond (e.g., Page and Shapiro 1983; Bartels 1991; Jacobs 1993; Erikson, Wright, and McIver 1993; Erikson, Mackuen, and Stimson 2002; Monroe 1979 and 1998; Wlezien 2004; but see
which groups govern? | 23
Page and Bouton 2006).14 However, citizens do not always agree with one another about what policies they wish their government to pursue. What kind of responsiveness does government provide when groups of citizens disagree with one another about what policy is best? As Dahl (1961) so concisely put it, “Who governs?” Political thinkers and practitioners have been particularly concerned with protecting the interests of numerical minorities when citizens disagree. Indeed, one of Madison’s chief justifications for the proposed constitution was that it would protect against majority tyranny: he claimed in Federalist No. 10 that “[t]o secure the public good and private rights against the danger of [a majority] faction . . . is then the great object to which our inquiries are directed” (Madison, Hamilton, and Jay 1987, 125). Almost two centuries later, Dahl (1967, 10) asked, “[H]ow can larger groups . . . be prevented from exploiting and tyrannizing smaller groups? . . . [W]ill there not be constant dissention, unrest, and even subversion by discontented minorities who find their aims thwarted by more populous groups?” Most studies that have examined various groups’ relative representation have been concerned with what Madison in Federalist No. 10 called minority, rather than majority, tyranny. Specifically, many have asked whether high-income groups exert undue influence on government. Early studies charged that in Washington and at the local level “moneyed interests” have the greatest impact on government decisions (Schattschneider 1960; see also Dahl 1961; Lowi 1969; Lindblom 1977; Polsby 1980). More recent research has found that the wealthiest Americans are much better represented than the poorest Americans in senators’ decisions and government policies (Bartels 2008; Gilens 2005; see also Hill and Leighley 1992). For instance, the American Political Science Association Task Force on Social Inequality recently documented the unequal representation of income groups via differential rates of participation, disparate responsiveness by government officials, and unequal policy outcomes (Jacobs and Skocpol 2005). These studies make an important contribution to our understanding of inequalities based on income, one of the important dividing lines in political representation that exists in the United States. Madison was less concerned about minority than majority tyranny, however, for, as he said in Federalist No. 10, “[I]f a faction consists of less than a majority, relief is supplied by the republican principle, which allows a majority to defeat its sinister views by a regular vote” (Madison, Hamilton, and 14. For a review of this literature, see Manza, Cook, and Page 2002.
24 | chapter two
Jay 1987, 125). Fewer studies have asked whether majorities wield too much political influence in the United States. Comparing the representation of African Americans, Latinos, and whites is a natural starting point for such an analysis, given the centrality of race in American politics (Hutchings and Valentino 2004) and the persistent disadvantages that African Americans and Latinos have faced in attempting to gain political equality as numerical minorities. Yet, there are large gaps in our knowledge of which racial or ethnic groups govern. We know relatively little about when and by how much minority groups are disadvantaged or advantaged compared to whites in procuring national policies they favor.
which groups govern? Scholars have documented the many obstacles to political equality that racial and ethnic minorities confront, such as their numerical minority status in most electoral districts, their disproportionate voting power based on patterns of party support (Bartels 1998), linguistic and citizenship barriers (De la Garza 2004), their generally lower rates of political participation and engagement (Verba, Schlozman, and Brady 1995), the infrequency with which these groups are targeted for mobilization (Rosenstone and Hansen 1993), and the dilution of their voting strength based on districting and electoral rules (Guinier 1994; Lijphart 1997), to name just a few. This state of affairs has led observers to worry that “blacks continue to be underrepresented in federal, state, and local government” (Guinier 1994, 8); there also is a widespread perception among leaders in the private and public sectors that African Americans occupy one of the lowest rungs of the political influence ladder (e.g., Verba and Orren 1985, 188–89).15 Similarly, others have concluded that there is “little or no direct or indirect substantive representation of Latinos” (Hero and Tolbert 1995, 648; but see Manzano and Norrander 2007). However, the studies of African American and Latino representation from which we have learned so much have tended to assume that whites are better represented than minority groups and proceeded to examine under what conditions African Americans and Latinos are better represented in absolute terms. Most studies of minority representation assess how the concentration of African Americans or Latinos in electoral districts shapes 15. There is some evidence that public officials have become more responsive to African Americans in recent decades (e.g., Keech 1968; Bullock 1981).
which groups govern? | 25
various legislative outcomes, including Democrats’ electoral success (Grofman, Griffin, and Glazer 1992), and legislators’ activities in office, typically their roll-call behavior (Combs, Hibbing, and Welch 1984; Grofman, Griffin, and Glazer 1992; Cameron, Epstein, and O’Halloran 1996; Lublin 1997; Canon 1999). These studies find that as the African American population in a district grows, representatives vote first more conservatively and then (with further growth) more liberally, although these patterns differ by region, district urbanicity, and legislators’ party affiliations and race or ethnicity (Combs, Hibbing, and Welch 1984; Grofman, Griffin, and Glazer 1992; Cameron, Epstein, and O’Halloran 1996; Canon 1999).16 Others have focused on public policy outcomes. For instance, state policies such as welfare spending are not more likely, and perhaps even less likely, to benefit minorities as their share of the state population increases (Wright 1977; Hero 1998; Johnson 2001). From our perspective, these studies of minority groups’ representation are absolutely vital, showing when and how minority groups’ absolute levels of representation rise. However, significant factors limit what we know about minority groups’ representation relative to that of whites. Minority representation studies nearly always compare the representation of African Americans or Latinos in some districts (e.g., districts with few minority constituents) to the representation of African Americans or Latinos in other districts (e.g., districts with many minority constituents).17 These investigations do not tell us whether or by how much African Americans or Latinos are under-represented compared to whites. Thus, they cannot say much about the political equality, or relative representation, of different groups. We must remember that African Americans and Latinos do not constitute monolithic political communities. There is significant variation in African Americans’ and especially Latinos’ political opinions (e.g., Gilliam 1996; Leal 2002; Tate 2003a; Gay 2004). As Gilliam (1996, 76– 77) puts it, “too much research on racial and ethnic minorities proceeds 16. In a similar vein, Bullock (1981) found that the advent of African American voting rights in the South after the 1965 Voting Rights Act produced a noticeable change in the tendency of southern representatives to support liberal positions and the roll-call alternatives supported by the Leadership Conference on Civil Rights (LCCR). 17. Exceptions include Hajnal, Gerber, and Louch (2002), who found that racial minorities were not much less likely than whites to prevail in the state initiative process, although they did tend to do less well on ballot proposals specifically dealing with racial policies, and Schumaker and Getter (1977), who report a bias toward the spending preferences of whites.
26 | chapter two
on the assumption of [attitudinal] homogeneity.” Latinos’ preferences are even more diverse, in part due to the different national origins of Latinos (Leal 2002). Thus, more liberal representatives hailing from districts with large minority populations may not be more representative of their African American or Latino constituents. To summarize, at this point in time, we do not know how large differences in representation are or whether these differences vary across issue domains. Moreover, we do not know how much these differences in representation are affected by such factors as the racial and ethnic composition of electoral districts, the race or ethnicity of elected officials, or political activity such as voting by minorities. We know, for example, that descriptively represented African Americans (those represented by members of their own group) are better represented than nondescriptively represented African Americans (Canon 1999; Lublin 1997). What we aim to determine is whether even descriptively represented minorities (or minorities who vote) remain underrepresented compared to whites, whether they become equally represented under these circumstances, or whether these conditions turn the tables on whites, leaving them less represented. Our research design, which focuses explicitly on the relative representation of groups, permits us to address these kinds of questions.
theoretical expectations From a theoretical perspective, why might some groups be better represented than others? When might the relative representation of groups change? We begin by assuming that elected officials desire reelection (Mayhew 1974). As many have pointed out, elected officials may have several goals, including serving the public, enacting certain public policies, and increasing their influence or prestige within their particular institution (e.g., Fenno 1973). However, meeting these goals requires maintaining elected office. Thus, reelection, although not necessarily the only goal, is a primary goal. Furthermore, in the particular context of citizen influence on government action, reelection is the main goal citizens can affect. As Fiorina (1974, 31) puts it, “[R]ealistically or cynically as the case may be, we believe that constituents’ preferences are reflected in a representative’s voting (if at all) primarily through his [or her] concern for his [or her] electoral survival.” If reelection is a primary goal, when making decisions elected officials will consider, perhaps among other things, how different actions will affect their reelection prospects. In the context of race and ethnicity in the United
which groups govern? | 27
States, elected officials are often forced to choose which groups to please. For example, if most whites are in favor of a policy and most Latinos are opposed, an elected official must please one group and disappoint the other.18 Of course, when different groups agree, an official can please them both. However, as plenty of opinion surveys demonstrate, African Americans, Latinos, and whites often disagree (e.g., Kinder and Sanders 1996; Uhlaner and Garcia 2002). We show this in detail in the next chapter. In the face of divided groups, we assume that the official seeking reelection will choose to please the group that will, in his or her mind, be most helpful in achieving that goal. Several factors shape a group’s ability to affect the probability of reelection. Group size is an obvious factor. All else being equal, especially turnout rates, larger groups can have more influence on reelection than can smaller groups. Second, differences in the salience of the issues at hand will shape the electoral consequences of actions. Various theoretical models of representative behavior have formally demonstrated the incentives for siding with a passionate minority over a passive majority (e.g., Downs 1957; Fiorina 1974). Consider an issue that Latinos are especially concerned with and whites care very little about. Pleasing whites, even if they are in the majority, will not generate much electoral benefit, while pleasing Latinos may reap electoral rewards. Empirical studies have shown that groups are more likely to pay attention to information in the media and election campaigns on issues they care more about and to make choices between candidates based on these salient issues (Iyengar 1990; Hutchings 2001 and 2003). Therefore, the group that cares more about the issue may have greater influence on officials’ reelection prospects. Consequently, officials may seek to please the passionate minority group when acting in this particular issue domain.19 Third, voting itself plays a role. Voters obviously affect election outcomes more than nonvoters do. Of course, office holders do not want to arouse a previously inactive opposition, but as Bartels (1998, 45) reminds us, “[V]ote-maximizing politicians must care more, other things being equal, about the views of regular voters than about the views of people who seldom or never get to the polls.” Empirical studies have added evidence 18. A third, more realistic, possibility, at least when a decision allows for alternatives, is to select a position that lies between those of the two racial groups. However, this risks offending both groups. In addition, the compromise, unless exactly in “the middle,” will still favor one group over the other. 19. The proliferation of interest organizations that monitor elected officials’ behavior in an issue domain and convey it to citizens in distilled form also facilitates this process.
28 | chapter two
to these contentions (e.g., Martin 2003; Griffin and Newman 2005; but see Ellis, Ura, and Ashley-Robinson 2006). So, the prevalence of voting within a group and the relative prevalence of voting across groups may affect the sensitivity of elected officials to the group’s interests. There are certainly other factors that affect the electoral success of government officials. However, these three factors are most obviously relevant to the relative importance of minority groups in reelection-minded officials’ decision making. If we take these three factors into account, we might expect to observe several empirical regularities. First, simply as a result of group size (not to mention greater resources), the views of whites will often prevail in government action. Nationally, whites comprise the majority of the population, which is also true in all but two states (where whites are a plurality) and a handful of congressional districts. Therefore, our baseline expectation is that when whites and minority groups disagree, whites’ preferences will win out. We might also expect that where African Americans or Latinos make up a larger share of an electoral district, they might be represented as well as whites are. However, this may be too simple a story. If the size of the minority population affects the similarity or dissimilarity of whites’ and minorities’ preferences, the effect of district racial and ethnic composition on minorities’ representation may not be perfectly linear. As it turns out, the gap between African American and white opinion does vary with the size of the African American population in congressional districts. According to the 2000 National Annenberg Election Survey (NAES), whites in districts with significant African American populations are somewhat more liberal than whites who reside in districts with very few African Americans, mostly because these tend to be more urban districts. For example, in districts where African Americans make up fewer than 10% of the constituency, 22% of whites identify themselves as liberal or very liberal. In districts with more than 40% African American constituents, 26% of whites called themselves liberal or very liberal. In contrast, African Americans become first more conservative, and then, with still further growth in their proportion of the population, they become more liberal, so that African Americans residing in districts with the largest proportion of African Americans are predicted to be only slightly more liberal than African Americans who live among few whites (see fig. 2.1). The important point is that empirical studies have found that as the African American population in a district grows, representatives vote first more conservatively and then (with further growth) more liberally. Other factors, too, may cause the relationship
which groups govern? | 29
figure 2 .1 . White and African American conservatism and district racial composition (2000 National Annenberg Election Survey).
between group size and group strength to be conditional (e.g., Combs, Hibbing, and Welch 1984; Grofman, Griffin, and Glazer 1992; Cameron, Epstein, and O’Halloran 1996; Canon 1999). Even where whites make up the majority, the other factors may on occasion overcome the power of numbers. In terms of salience, we would expect that if whites and African Americans or Latinos care equally about an issue (or whites care more), then whites will carry the day because of greater numbers. However, if Latinos or African Americans care more than whites about an issue, the balance of representation may begin to even out. Thus, we would expect that on issues that African Americans or Latinos care about more than do whites, these groups’ representation will be more on par with that of whites. As we will see in chapter 3, African Americans as a group tend to be more concerned than whites about policies related to welfare, health care, and crime, while Latinos tend to be more concerned than whites about welfare, education, crime, and health care. Finally, we might expect that increased turnout among minority groups would boost their representation relative to that of whites. After all, studies have found that voting generally has its rewards (e.g., Campbell 2003; Martin 2003; Griffin and Newman 2005). Historically and up to the present
30 | chapter two
day, minorities have voted at rates that do not match those of whites. So, boosting turnout only among minorities may improve these groups’ relative representation. However, as we will see in chapter 8, some votes might bring more rewards than others. If so, increasing turnout evenly among all groups, including whites, may not yield equality of representation among these groups.
contributions Armed with these theoretical expectations, we have designed our study to answer important questions about the political equality of minority groups in American politics. First, we take stock of the extent to which whites are better represented than African Americans and Latinos. This shifts the analysis of minority representation, traditionally concerned with the representation of one particular group in isolation, more explicitly toward questions of political equality. This is significant because people appear to care deeply about their standing relative to others. As Robert Dahl (2006) puts it, [H]uman beings are naturally endowed with a sensitivity to the unequal distribution of rewards to others whom they view as comparable to themselves in relevant ways. Whether . . . we describe this sensitivity by the aseptic term inequity aversion, or use earthier language like jealousy or envy, what a human being sees as unfairness or injustice will often arouse strong emotions. Given the opportunity, these emotions will then express themselves in actions, which may range from an immediate verbal expression— “It’s not fair”—to behavior intended to bring about a fairer distribution, whether by peaceful persuasion or violence. (39)
Members of minority groups are no strangers to these sentiments, which Dahl argues manifest a universal concern for equal distribution of anything of value. In fact, minority group members may even care more about their relative standing since, historically, it has generally not been as favorable as that of whites. Recall from chapter 1 that the NAACP’s first objective, according to its mission statement, is to ensure “the political, educational, social and economic equality of all citizens.” These goals are explicitly relative—comparing African Americans’ political, educational, social, and economic status to that of whites. Moreover, every two years from 1984 to 2000 (except 1998) the American National Election Studies have asked Americans
which groups govern? | 31
whether they agreed with the statement, “The country would be better off if we worried less about how equal people are.” Over this period, 52% of whites agreed or strongly agreed with this statement, while 46% of Latinos and just 38% of African Americans did so. In addition, experimental results from the “ultimatum game,” in which one player first proposes a split of a sum of money and a second player decides whether to accept the proposal or for both players to receive nothing, suggest that African Americans care deeply about inequality (Eckel and Grossman 2001). For instance, one study found that African American “proposers” averaged a $2.11 offer when splitting $5.00, while non–African Americans’ average offer was $1.79, suggesting that African Americans generally offered a more even split than did others. Even more tellingly, African Americans were much more likely to reject unfair proposals: African Americans rejected nearly four times as many $1.50 offers as did non–African Americans (47% to 13%). Nearly three in four African Americans rejected $1.00 offers, compared to less than half of non–African Americans. Given a choice between the status quo and an alternative state of the world in which they are better off in absolute terms but worse off in relative terms, a preponderance of African Americans (but not whites) are much more likely to opt for the status quo.20 Students of democracy are also concerned with both absolute and relative representation. Dahl (2006, ix) contends that “the existence of political equality is a fundamental premise of democracy.” Of course, we want to know when any group’s absolute representation increases or decreases. This tells us about the government’s responsiveness to public opinion, a cornerstone of democratic politics. However, democratic concern for equality also highlights the importance of knowing if and when some groups are better represented than others. As we have noted, many extant studies examine the former question. Our study examines the latter. Second, in a related way, we explore whether and why groups are sometimes politically equal. Where earlier studies demonstrate that the racial and ethnic composition of districts or descriptive representation tend to improve African Americans’ or Latinos’ representation, we ask whether these gains lead to political equality, or if these gains are so small relative
20. Other experimental games show that racial minorities are more averse than whites to the risk of inequality—that minorities prefer a modest, equal improvement in the welfare of all the participants in the game to the risk that all participants may receive either a more generous, equal benefit or a more generous (in the aggregate) but unequal benefit (Hong and Bohnet 2004).
32 | chapter two
to the size of the inequality that they leave African Americans and Latinos unequally represented compared to whites. This enables us to see whether there are some conditions under which minority citizens are currently equally represented compared to whites. Highlighting the relative representation effects of factors like descriptive representation and issue salience also suggests reforms that might lead toward greater political equality. Third, we ask if any of the three standards of political equality we have articulated are met. We provide empirical evidence to assess whether minority group representation is more than proportional, as advocates of race-conscious egalitarianism would require. We also determine whether minorities are equally represented as individuals, as a proportional standard of equality demands. For example, we measure citizens’ opinions directly and explore how district racial and ethnic composition amplifies or dampens the connection between these opinions and representatives’ voting decisions. We also compare the representation of individuals where groups are of roughly equal size. Finally, we ask whether minorities are more than proportionally represented on issues that they care more about than whites, consistent with a pluralist view of democracy. Our articulation of multiple standards of equality advances our understanding of minority representation by allowing us to identify more precisely the circumstances under which minorities are politically equal. Fourth, we contribute to the study of political equality by focusing on the equality of national government outputs. Such a focus is relatively novel. As Sidney Verba observes, “[T]he literature on the receipt of messages and the [equal] response to them is not as well developed as that on the [equality of the] messages sent” (2003, 666). Moreover, studying outputs is important. As Polsby (1980, 484) emphasizes, “[R]esearchers should study the outcomes of actual decisions . . . and not conclude prematurely that the combination of intentions and resources inflexibly predetermines outcomes.” To be sure, several studies of racial differences in policy responsiveness examine state policymaking (Hero 1998; Manzano and Norrander 2007). Many of the policies that have important effects on African Americans’ lives, such as policies concerning spending on education and welfare eligibility, are made at the state level (e.g., Hero 1998). Our focus on national policies is important because the federal government is increasingly viewed by minorities as an important vehicle for advancing their public interests (Dawson 1994). Relatedly, many national policies have a large impact on minorities’ lives, even if some federal programs are ad-
which groups govern? | 33
ministered by state governments. Last, studying groups’ representation at the national level permits us to measure African Americans’ and Latinos’ opinions directly, rather than using demographic proxies of opinion as prior studies have done. Finally, we take into account both citizens’ income and their race/ ethnicity. By examining race and ethnicity together with income, we advance the study of political equality beyond its typical focus on income alone (e.g., Bartels 2008; Gilens 2005; Jacobs and Skocpol 2005; Lardner and Smith 2005). One reason that several studies examine the relative representation of income groups but few explore racial and ethnic groups in this way may be the presumed equivalence of race/ethnicity and income. These often go hand in hand. For instance, where we have emphasized the racial composition of September 11th and Katrina victims, others have focused on the lofty incomes of many September 11th victims, and the deep poverty “exposed” by the Katrina disaster. After all, 14% of the September 11th victims earned more than $200,000 per year. Meanwhile, according to Census-based calculations by the Center on Budget and Policy Priorities, approximately one in five residents in the counties declared federal disaster areas in Louisiana, Mississippi, and Alabama after Katrina were living in poverty before the hurricane struck.21 Although race and income are undeniably related, race is hardly a perfect predictor of income. For instance, Radcliff and Saiz (1995, 776) argue that class “hardly captures the distinctive differences between whites and blacks.” Indeed, the gap between the median incomes of whites and African Americans appears to have shrunk over time. Using Census data, figure 2.2 plots the difference in the median incomes of whites and African Americans and whites and Latinos over time. According to the figure, the gap between African Americans’ and whites’ median incomes has declined approximately 50% since 1960. There is still a considerable difference between the incomes of whites and African Americans, but this disparity has declined substantially. One reason for this is the well-documented emergence of a sizeable African American middle class in recent decades (see, e.g., Dawson 1994). On the other hand, figure 2.2 shows that the income gap between non-Latino whites and Latinos has increased in recent years. One possible (but contentious) reason for this trend is the large-scale legal and
21. See the Center on Budget and Policy Priorities Web site: http://www.cbpp.org/9-1905pov.htm.
34 | chapter two
figure 2 .2 . Group differences in income, 1948–2003 (U.S. Census).
undocumented immigration of low-income Latinos to the United States over the past twenty years.22 In sum, while racial and ethnic differences in income are significant, race is an imperfect predictor of income and increasingly so for African Americans. Although income plays a significant role in shaping politics in the United States, it is important that studies of income differences in political representation be supplemented with studies of racial and ethnic differences in representation. To be sure that our race- and ethnicity-based findings are not merely masking income-based inequalities, wherever possible we control for income in our analyses. In the end, we show that both race/ethnicity and class are related to political inequality. Having defined the critical concepts and outlined our expectations, we are now ready to focus on the political preferences and priorities of Latinos, African Americans, and whites. The next chapter demonstrates the extent
22. Increases in legal immigration stem from the 1986 Immigration and Control Act, which offered amnesty to undocumented aliens, and the Immigration Act of 1990, which increased annual immigration quotas. The U.S. Citizenship and Immigration Services (USCIS) also estimates that 150,000 undocumented aliens per year enter the United States from Mexico, and that more than 3.5 million citizens of Mexico and Central and South American countries reside in the United States.
which groups govern? | 35
to which these three groups make different demands of government and care about different issues. Identifying differences in policy preferences and priorities enables us to define more substantively the issue domains in which we expect to see greater or lesser degrees of inequality in political outcomes, as measured against the three standards of equality described in this chapter.
part two
Differences in Representation In this part of the book, we establish a baseline answer to the “Who governs?” question. For one group to be better represented than another, the groups first must hold different preferences about the actions they wish government to take. Otherwise, even if the decisions of government are only intended to reflect the preferences of one of the groups, substantively both groups would enjoy the same responsiveness from government. Moreover, it would be impossible for us to know in this case to which group government was being responsive. Thus, our first task is to establish that African Americans, Latinos, and whites have different policy preferences and different policy priorities. In part 3, these analyses will guide our expectations concerning policy domains where minorities might better represented because of issue salience. We then get to the heart of the matter, asking whether groups’ preferences for policy change are equally represented in the policy outputs of government. We also test whether African Americans, Latinos, and whites are equally represented in the decisions of members of Congress, judged against various standards of equality.
th r e e
Differences in Political Preferences and Priorities Race has divided American society and politics from the beginning. Keen political observers have long feared that racial divisions were unbridgeable and would ultimately rend the republic at its core. Tocqueville considered race relations “the most formidable of all the ills that threaten the future of the Union” and believed emancipation would lead to a race war (Kinder and Sanders 1996, 11). Thomas Jefferson was even less hopeful, finding a harmonious racial society inconceivable. Rather, he predicted in Notes on the State of Virginia (1955, 138) that emancipation of slaves and the establishment of a biracial society would lead to “convulsions which will probably never end but in the extermination of one or the other race.” Even Abraham Lincoln considered the establishment of black colonies in the Caribbean or Central America the best solution to racial tensions. As he argued in a debate with Stephen Douglas, blacks’ and whites’ physical differences were so great that they “forever forbid the two races living together upon the footing of perfect equality” (1953, 3:16). Even after the Emancipation Proclamation, the passage of the Thirteenth, Fourteenth, and Fifteenth Amendments, and the Civil Rights Movement, race continues to divide American society and politics. In the late 1960s, as race riots raged, Otto Kerner, governor of Illinois, and his National Commission on Civil Disorders warned that the United States was “moving toward two societies, one black, one white—separate and unequal” (National Advisory Commission on Civil Disorders 1968, 1). More recently, Stephen Carter claimed that “[r]ace, more than any other organizing category . . . continues to drive our nation into frenzies that lead to unparalleled viciousness” (1994, xvii–xviii). These divisions are a defining feature of American politics. In fact, Kinder and Sanders (1996, 27) conclude from years of opinion surveys that the differences in policy preferences between
40 | chapter three
African Americans and whites is “a divide without peer.” Over the past decades, the dividing lines have changed from simple black and white characterizations to include several growing racial and ethnic groups, most notably Latinos. As we will see, both in terms of broad political perspectives and on specific issues, the views of African Americans, Latinos, and whites are quite distinctive. In this chapter, we introduce the data we use to measure individuals’ political preferences and then describe the differences in the political preferences and priorities of African Americans, Latinos, and whites in several issue areas. We find that the differences in groups’ preferences vary greatly from issue to issue. Moreover, while there are many similarities between African Americans’ and Latinos’ preferences relative to those of whites, we see that it is important to analyze these groups separately. Beyond differences in policy preferences, we note that the minority groups have distinctive policy priorities as well. Not only do African Americans, Latinos, and whites want the government to do different things, the issues these groups care most about differ. These preference and priority differences lay the groundwork for the remaining chapters in part 2. Chapter 4 compares the groups’ preferences with the policies of government and chapter 5 compares the decisions of individual government actors.
data sources Our chief source of opinion data is the 2000 National Annenberg Election Survey (NAES), of which we use portions that include a total of 57,197 respondents.1 These surveys include more than 5,000 African American and 5,000 Latino respondents, a sample that few mass-level political surveys can match. The sheer size of the NAES enables us to overcome some of the technical hurdles to examining critical substantive questions. For example, the large sample allows us to estimate the preferences of the three groups in many states. Most surveys simply do not include enough respondents in a substantial number of states to allow reliable state-level analyses. Additionally, unlike approaches using other surveys (Erikson, Wright, and McIver
1. To generate our data set, we combine the NAES national cross-section data, which were obtained between December 1999 and January 2001, the NAES Super Tuesday cross-section, and the NAES Second Tuesday cross-section. For the sampling protocol of the NAES, see Romer et al. (2004).
differences in political preferences and priorities | 41
1993; Brace et al. 2002), the NAES does not need to be aggregated over a long period of time (just over a year) to generate state-level estimates. Pooling over longer periods raises concerns about aggregating data that may not be stable over time. We also rely on other data sources for measures of political preferences among the three groups. In particular, the General Social Survey (GSS) conducted by the National Opinion Research Center (NORC) regularly asks respondents about their preferences for national government spending in various areas. We examine these spending preferences for the period 1973– 2002. In addition, we rely on data from the National Election Studies (NES) to compare the groups’ policy priorities and for several other purposes. These surveys were conducted from September to November of even-numbered years, 1974–2002.2
differences in ideology We begin broadly, by comparing the general ideological orientations of the three groups, asking how liberal or conservative they tend to be. In the nation as a whole, these groups hold distinctive ideological views. The NAES asked respondents to place themselves on a five-point ideology scale ranging from very liberal (recoded 1) to very conservative (recoded 5).3 A straightforward way to compare the groups’ ideological orientations is to examine the distribution of self-placements for each group. As figure 3.1 shows, whites are the most conservative of the three groups, with an average ideology of 3.19, while African Americans are the most liberal group with a mean of 3.01, and Latinos are roughly in between with a mean of 3.10. In each case, the differences between whites and minorities are statistically significant. In terms of the distribution across the five categories, although approximately equal proportions of each group claim to be very conservative, whites are much more likely than African Americans to be conservative, and somewhat more likely than Latinos to be conservative. Meanwhile, a greater proportion of African Americans identify themselves as very liberal, liberal, and moderate than either of the other two groups. To convey the magnitude of
2. We also use NES data to measure racial groups’ spending preferences in dealing with crime, as no comparable item was included in the NAES. 3. The wording of the ideological orientation question was, “Generally speaking, would you describe your political views as very conservative, conservative, moderate, liberal or very liberal?”
42 | chapter three
figure 3 .1 . Ideological orientation, African Americans, Latinos, and whites (2000 National Annenberg Election Survey).
these differences, we note that the difference between the mean ideology of whites and African Americans (.18) is about three times the size of the difference between the average ideology of NAES respondents with incomes between $15,000 and $25,000, the 25th percentile in the income distribution, and NAES respondents with incomes between $50,000 and $75,000, the 75th percentile in the income distribution. Finally, we note that using the NAES five-point ideology scale rather than the seven-point scale used in other surveys may minimize the reality of group differences in ideology.4 We also note that the variance in all three groups’ ideologies is similar (with minorities actually exhibiting slightly more variation than whites).5 4. For instance, 2002 NES data indicate that more than four times as many African Americans as whites identify themselves as “extremely liberal,” and nearly twice as many African Americans as whites identify themselves as “liberal.” Pooling NES data from 1990 to 2002 to compare whites and Latinos reveals that even using a seven-point ideological scale, Latinos are modestly more liberal than whites. For instance, about 60% of Latinos identify themselves as either extremely liberal, liberal, slightly liberal, liberal, or moderate, compared to about 55% of whites. Unfortunately, the much smaller sample size of the NES does not permit us to use this measure in our analyses, but the greater liberalism of African Americans in the NES suggests that our results may somewhat understate racial differences in representation. 5. The standard deviation of white ideology is 0.93, while the standard deviations of African American and Latino ideology are 0.97 and 0.98, respectively.
differences in political preferences and priorities | 43
While there is certainly some truth to the conventional wisdom that African Americans are more politically cohesive than whites, overwhelmingly supporting the Democratic Party and its candidates (Frymer 1999), there clearly is significant variation among African Americans’ (and Latinos) ideological orientations (see also Dawson 1994; Gilliam 1996; Canon 1999; Uhlaner and Garcia 2002; Tate 2003a). Similarly, more Latinos support Democrats than Republicans, but these tendencies clearly do not indicate uniformly liberal ideologies among Latinos. This point has substantive and technical significance. Substantively, if African Americans or Latinos as a group hold a diversity of ideological orientations, simply electing liberal officials (usually Democrats) may not lead to improved representation for all or even most members of minority groups. In addition, it is generally easier to represent a more homogeneous group compared to a group with heterogeneous preferences simply because it is impossible to please everyone (or even a sizeable majority) in a diverse group. Thus, diversity in all three groups poses a representation challenge for any elected representative. More technically, significant ideological variation is important because without sufficient variation within a group’s preferences, it is difficult to observe some types of representation as commonly measured by scholars of Congress (e.g., Miller and Stokes 1963; Achen 1978; Bartels 1991; Ansolabehere, Snyder, and Stewart 2001). It is possible that these ideological differences are driven by large differences in group preferences in just a few states. However, when we look at the state level, we continue to find widespread and significant differences. We must first confront the difficulty that there are relatively few African Americans and Latinos in many states. Fortunately, the NAES provides a sample with at least 50 African Americans and an average of more than 200 African Americans sampled in 26 states, and at least 48 Latinos and an average of more than 250 Latinos in 21 states. These samples provide reasonably valid and reliable estimates of state-level opinions for the three groups, although predictably, the measures were somewhat less reliable for the minority groups.6 As we proceed with our analysis, we pay particular 6. Under standard definitions of reliability (see Jones and Norrander 1996), state-level mean white attitudes were highly reliable, while African Americans’ and Latinos’ state-level ideological orientations were less reliable. The reliability coefficient ranges from 0 to 1. Jones and Norrander classify values over .7 as “highly reliable,” between .6 and .7 “moderately reliable,” and below .6 as “unreliable” (302). The reliability coefficient for white ideology was .97; for African American and Latino ideology, it was .53 and .44, respectively. These state-level ideology measures are highly correlated with the percentage of the 2000 presidential vote George Bush received in the state minus
44 | chapter three
figure 3 .2 . Bias in state ideology (2000 National Annenberg Election Survey).
attention to the relatively less precise measures of minority group opinion and compensate accordingly. For now, the important point is that, among NAES respondents, there are significant differences at the state level between whites and African Americans and between whites and Latinos.7 Figure 3.2 presents the difference between African Americans’ and whites’ ideology in each state (see black columns) along with the difference between Latinos’ and whites’ ideology in the relevant states (see white columns). Positive values in figure 3.2 indicate the percentage Al Gore received (r = .93), and Erikson Wright, and McIver’s (1993) state ideology measure (r = .94). As an additional validity check, we correlated our measures of white ideology and African American ideology with Erikson, Wright, and McIver’s (subsequently, EWM) pooled CBS/New York Times data (1978–93). The two measures of white ideology were highly correlated (both .94), while the two measures of African American ideology were somewhat less highly correlated (.58). This may be attributable to the changing attitudinal composition of the African American population (e.g., Dawson 1994; Tate 2003a), given that our data begins more than two decades after EWM’s data begins. Furthermore, the geography of state white, African American, and Latino ideology also comports with intuition. For instance, African American ideology is more conservative in the South (FL, LA, NC) and more liberal in the Northeast (MA, NJ, NY) and on the Pacific Coast (CA). Latino ideology is more conservative in Florida and Texas than it is in California. 7. For more on racial differences in state-level preferences, see Norrander (2005).
differences in political preferences and priorities | 45
that the mean white opinion in a state is more conservative than the mean African American or Latino opinion. As a group, African Americans are significantly more liberal than whites in their general orientation toward politics, and in most states this difference is most likely not due to chance alone.8 In contrast, the difference between Latinos’ and whites’ ideologies within states is generally quite a bit smaller.9 Like African Americans, Latinos are more liberal in the majority of states where Latinos and whites exhibit different ideologies. However, in several Northeastern states whites are more conservative than African Americans yet more liberal than Latinos (NY, MA, PA). In about half the states we examine, the difference between white and Latino opinion is statistically significant.10 The noticeable variation in the difference between whites’ and African Americans’ ideologies and whites’ and Latinos’ ideologies across states begs for explanation. For instance, differences in white and African American ideology appear to be larger in the South than elsewhere. Why states vary in the difference between whites’ and minorities’ attitudes is an interesting and important question, and insofar as we know, unanswered. However, the principal goal of this study is to examine whether variation in these differences in attitudes relates to the manner in which elected officials vote on public policies, and so for our purposes it is sufficient to demonstrate that these differences exist. There is plenty of ideological disagreement between whites and African Americans and between whites and Latinos within states, and thus the possibility of disparate representation across the full range of issues decided by statewide elected officials. It is important to note a potential pitfall of our use of ideological orientation to measure minorities’ general preferences across many policies. Ideological orientation is generally believed to relate less strongly to some political characteristics such as party identification among Latinos and African Americans as compared to whites (Uhlaner and Garcia 2002, 98;
8. p < .10 in 20 of 26 states (two-tailed test). We also compared white and African American ideology using Erikson, Wright, and McIver’s pooled CBS/New York Times measures of ideology (1976–93), which were measured on a three-point scale. We found that among the 33 states that contained more than 50 African Americans in the EWM data, 27 states show a significant difference between white and African American ideology (p < .10). 9. This is not solely driven by Cubans, who tend to be more conservative (Uhlaner and Garcia 2002). Excluding Cubans does not significantly alter the mean for Latino opinion, mostly because there are few Cubans in the NAES sample (191). 10. That is, in 10 of 21 states p < .10, two-tailed test. The states with significant differences are CO, GA, MI, NC, NM, OH, PA, TX, VA, and WA.
46 | chapter three
McClain and Stewart 2002). However, ideological orientation relates similarly to many of the issue positions of whites, African Americans (Tate 2003a), and Latinos (Uhlaner, Gray, and Garcia 2000; Uhlaner and Garcia 2002). According to Tate (1993, 31–32), while a substantial proportion of African Americans do not offer an ideological orientation when queried, so that liberal-conservative ideology may not be a useful construct for this group, African Americans who do identify themselves as liberal or conservative tend to converge on a common set of reference points for this self-assessment. Namely, like whites who identify themselves as liberal, African Americans who identify themselves as liberal favor equality and social reform, including a commitment to investment in social programs. Similarly, scholars have concluded that ideology “is useful in understanding Latino opinions,” even if its usefulness is limited (Uhlaner and Garcia 2002, 99). However, the concern that ideology may be a poorer measure of preferences for minorities is one of several reasons to go beyond general ideological orientations and compare the groups’ preferences on specific issues.
african americans’ and whites’ preferences on specific issues Even at the level of specific issues, these three groups exhibit different preferences. We first compare African Americans’ and whites’ preferences on specific issues; then we compare Latinos’ and whites’ preferences. Many scholars have examined the specific issue preferences of African Americans and whites, and in many issue areas, these preferences diverge considerably (Tate 1993; Kinder and Sanders 1996; Canon 1999; Kinder and Winter 2001). These groups’ mean attitudes differ dramatically on issues directly related to race, such as affirmative action, job discrimination, and civil rights, and somewhat less so on “implicitly racial” issues like education, health care, and welfare spending (Kinder and Sanders 1996; Lublin 1997; Canon 1999). These differences in attitudes may arise from African Americans’ “distinctive, and difficult” place in the nation’s history (Canon 1999, 21), from “group animosities and solidarities” (Kinder and Sanders 1996, 7; Kinder and Sears 1981), from differences in political values and principles (Sniderman and Piazza 1993; Kinder and Winter 2001), and from the ways elites frame issues (Kinder and Sanders 1996). We need not concern ourselves with the source of these differences in political preferences—it is useful for our analysis simply to know that they are widely agreed to exist. On the other hand, African Americans’ and whites’ attitudes are believed to differ much less on issues that have
differences in political preferences and priorities | 47
little racial content, such as foreign affairs, immigration, and social issues such as abortion and school prayer (Kinder and Sanders 1996; Canon 1999). We examine differences in preferences for federal government spending on various types of programs. These include implicitly racial issues such as welfare, education, health care, and crime, as well as potentially race-related issues such as the environment and defense.11 To make our comparisons, we generally use the 2000 NAES, but for one issue, we draw on comparable NES data.12 Respondents were asked if the federal government should spend “more, the same, less, or none” in these domains. Using this data, we estimated an ordinary least squares (OLS) model with race (0 white, 1 African American) predicting preferences for increased spending on social security, education, health care insurance, Medicare, aid to mothers with small children, crime, defense, and the environment.13 From these estimates, we calculated the predicted level of support for increasing spending. If a group’s predicted response equals four on this measure, all members of the group favored more spending; if three, the average member of the group favored no change in spending, if two, the average member favored less spending, and if one, no spending at all. On six of eight issues, we found that African Americans were more likely than whites to prefer more spending (see fig. 3.3; compare the solid lines).
11. Several recent studies have documented instances of “environmental racism”: situations that generate racial differences in environmental quality (Camacho 1998; Adamson, Evans, and Stein 2001). Because 21% of U.S. military volunteers are African American, defense spending may have special consequences for minority groups due to their overrepresentation in the armed forces (Fears 2003). 12. The NAES question wording was as follows: “Social Security benefits—should the federal government spend more money, on this, the same as now, less, or no money at all?” Responses are coded from 1 (“no money at all”) to 4 (“more money”). The remaining lead-ins were worded as follows: “Providing financial assistance to public elementary and secondary schools”; “Providing health care for people who do not already have it”; “Providing health care for elderly people, usually called Medicare”; “Providing assistance to poor mothers with young children”; “Maintaining a strong military defense”; and “Protecting the environment and natural resources.” The environmental item asked if the government should expend more or less effort rather than resources. We use the NES surveys from 1984, 1992, 1994, 1996, 2000, and 2002 to measure preferences for dealing with crime. These use a three-point scale, asking respondents whether spending should be increased, kept the same, or decreased/ cut out entirely. For comparability with the NAES items, we coded these 4, 3, and 2 respectively. 13. Ordered probit models yield very similar results, but are more difficult to display graphically.
48 | chapter three
figure 3 .3 . Predicted support for increasing spending in eight policy areas, whites and African Americans (2000 National Annenberg Election Survey).
The only issues we examined for which whites and African Americans showed similar support for spending were military spending and spending to deal with crime, with whites a little more supportive of spending on the former and African Americans a little more supportive on the latter. In terms of absolute rather than relative levels of support, the average African American was quite near to supporting an increase in spending in five of the eight issue areas, with somewhat less enthusiasm for spending to deal with crime, even less but still substantial support for welfare, and relatively little support for military spending. By comparison, the average white respondent was quite ambivalent about increasing spending in seven of the eight issue areas, with a group mean near 3.5, while the average white favored no increase in spending to help mothers with young children. We also examined a separate, implicitly racial NAES item related to federal government effort rather than spending per se: whether the respondent believed that the government should or should not try to reduce income differences between rich and poor Americans. We found that 69% of African Americans thought that this should be a goal of government, while only 46% of whites thought so. Notably, where differences in political attitudes exist, they do not appear to be reducible to differences in income. Indeed, one study found that
differences in political preferences and priorities | 49
the estimated size of the racial gap in opinion between whites and African Americans on explicitly racial policies such as fair employment, federal spending on programs that assist African Americans, and the preferential hiring of African Americans was not reduced at all after accounting for family income, level of educational attainment, and occupational status (Kinder and Sanders 1996, 299; see also Kinder and Winter 2001). In the NAES, we similarly found that African Americans’ greater liberalism regarding the government’s proper level of involvement in preventing job discrimination against African Americans as well as whether the Confederate flag should be removed from the South Carolina state capitol building were robust to a control for income.14 When we control for income in the models of spending preferences, the racial divide remains vast, as the dashed lines in figure 3.3 indicate. The differences in preferences among lower-income whites and African Americans were only slightly attenuated, so racial differences in spending preferences are not simply proxying differences in income groups’ preferences. The same is true for preferences about government attempts to reduce income inequality. Among whites and African Americans earning less than $35,000 per year, 74% of African Americans agreed that this should be a goal of government, compared to 57% of whites. In sum, even on many implicitly racial issues such as federal spending on welfare, health insurance, and education, African Americans and whites who share the same income express rather distinctive preferences, creating the possibility for unequal representation on these issues. These patterns of spending preferences are also found in the GSS and Roper surveys that we use in later analyses. Beginning in 1972, in most years the GSS has queried citizens about their preferences for government spending in nine issue areas useful for our analysis: national defense, the environment, education, crime prevention, health care, foreign aid, welfare, aid to cities, and the space program. Respondents were asked, “We are faced with many problems in this country, none of which can be solved easily or inexpensively. I’m going to name some of these problems, and for each one I’d like you to tell me whether you think we’re spending too much money on it, 14. In an ordered probit model of the job-discrimination item, the parameter estimate for an African American indicator was 1.31 in a bivariate model. Controlling for income, the estimate for the African American indicator fell just to 1.25. In a probit model of the Confederate flag item, the parameter estimate for an African American indicator was 0.78 in a bivariate model, and actually rose to 0.90 in a model controlling for income.
50 | chapter three
figure 3 .4 . State-level bias on race-related issues (2000 National Annenberg Election Survey).
too little money, or about the right amount.”15 Unfortunately, the GSS did not go to great effort to distinguish Latino respondents from the “other” category until 2000, so we rely exclusively on the Roper surveys in our analyses of Latinos. In the GSS/Roper surveys, African Americans continue to favor increased spending more than do whites. Latinos are again closer to whites, but tend to favor increased spending to a greater degree than do whites. These national-level differences in specific issue preferences extend to the states as well. We generated a state-level measure of opinion on these kinds of racial and implicitly racial issues for whites and African Americans. Using principal components factor analysis, we combined responses to two NAES items concerning race-related issues: whether the federal government should do more to end job discrimination against African Americans, and whether poverty is a serious national problem.16 Figure 15. This statement was followed by these phrases: “space exploration program,” “halting the rising crime rate,” “improving and protecting the environment,” “improving and protecting the nation’s health,” “solving the problems of the big cities,” “improving the nation’s education system,” “the military, armaments, and defense,” “foreign aid,” and “welfare.” 16. The exact item wordings follow: “Trying to stop job discrimination against blacks—should the federal government do more about this, the same as now, less, or
differences in political preferences and priorities | 51
3.4 presents the racial differences in citizens’ “racial ideologies” across the twenty-six states for which we have already observed differences in whites’ and African Americans’ overall ideological orientations. Again, positive scores indicate that whites are more conservative than African Americans. In every state, the mean white opinion on these issues is more conservative than the mean African American opinion, and in all twenty-six states the difference in means is statistically significant (p < .01; two-tailed test). There is also some variation in the racial bias of opinion on these issues, with several southern states (LA, GA, MS, AL) exhibiting some of the greatest differences between whites and African Americans, and several northeastern states (MA, NY) exhibiting some of the smallest differences.
latinos’ and whites’ preferences on specific issues Next, we compare the preferences of Latinos and whites on specific issues. Before doing so, however, we confront the reality that “Latino opinion” is something of a misnomer, because the Mexican, Puerto Rican, Cuban, and other Latino populations that comprise this group often differ markedly in their attitudes (e.g., De la Garza et al. 1992; Leal 2002). Rather than an obstacle to measuring differences in political representation, however, this as an analytical advantage. If we wish to observe whether senators are responsive to variation in Latino attitudes across the country, that is, whether states with conservative Latinos have more conservative senators than states with more liberal Latinos, there must be variation in Latino attitudes. Moreover, there is considerable variation in whites’ and African Americans’ preferences, yet we do not hesitate to define these as politically relevant groups. There appear to be “more differences than similarities in the opinions of Anglos and Latinos” (Leal 2002, 33; see also Uhlaner and Garcia 2002). In general, prior studies find that Latinos are somewhat more liberal than whites on a variety of policies, including bilingual education and affirmative action (Leal 2002; De la Garza et al. 1992; Uhlaner and Garcia 2002). We nothing at all?”; and “The amount of poverty in the United States—is this an extremely serious problem, serious, not too serious, or not serious at all?” This analysis retained one principal factor (eigenvalue = 1.54), with factor loadings of 0.56 and 0.57 for the two items. We then generated a factor score for each NAES respondent (both white and African American) answering both items, and used the state white and African American means of these scores to measure racial groups’ state-level preferences on race and race-related issues.
52 | chapter three
extend these investigations to the same issue domains in which we compared whites’ and African Americans’ preferences above. To compare the gap in white/Latino opinion on specific issues to that in white/African American opinion and to determine the extent to which differences in white/Latino opinion are explained by differences in income, we return to questions about government spending in the eight issue domains used above, this time comparing whites and Latinos. As figure 3.5 shows, Latinos are quite a bit more favorable than whites toward increased spending for education and welfare aid to mothers with dependent children (compare solid lines). There are other issue differences, but the gaps between the two groups are not as great as we observed between whites and African Americans in figure 3.3. Also, in contrast to what we observed in figure 3.3, Latinos and whites have a somewhat greater difference in preferences on spending to deal with crime. Finally, income appears to account for somewhat more of the gap in white/Latino opinion than the gap in white/African American opinion (see dashed lines). Still, there are clear differences in white and Latino opinion, even within income groups. As we did above, we also examined whether white or Latino respondents were more likely to believe that the government should try to reduce income differences between rich and poor Americans. We found that 68% of Latinos thought that this should be a goal of government, while only 46% of
figure 3 .5 . Predicted support for increasing spending in eight policy areas, whites and Latinos (2000 National Annenberg Election Survey).
differences in political preferences and priorities | 53
whites thought so; this difference in opinion was modestly smaller among lower-income Latinos and whites (74% to 57%, respectively). Finally, we compared white and Latino opinion in one additional area—immigration. It appears that Latinos are only a little more liberal than whites on immigration policy, with nearly identical proportions of whites and Latinos indicating that the rate of immigration to the United States is an extremely serious problem (27%) and only about 4% more Latinos than whites indicating that immigration is not a problem at all. This is consistent with what prior studies have found (Uhlaner and Garcia 2002). To sum up, we have seen that in their general ideological orientations, African Americans are more liberal than whites. This is true across the nation as a whole, as well as within at least half of the nation’s states. These differences extend to preferences for government action on specific issues. African Americans and whites differ most on spending for welfare, health care, education, and the environment, with smaller differences on defense spending. In some issue areas, both whites and African Americans prefer more spending, but as a group African Americans are more in favor of increased spending than whites, while in other areas, only African Americans prefer more spending. Latinos are more liberal in their general ideological outlooks than whites, but less liberal than African Americans. Latinos differ most from whites in their attitudes toward spending on education, spending to reduce income differences, and spending to deal with crime. Finally, minorities’ distinctive preferences are not explained away by income differences across groups. These differences in many of the issue positions create the possibility that some groups are better represented in the decisions of government, and that we can uncover evidence of this disparate representation in some issue domains.
differences in issue priorities In the areas where we have observed differences in the issue positions of these groups, we will also want to know if there are differences in their issue priorities. To answer this question, we must identify issues that are of higher priority for African Americans and Latinos, that is, issues that these groups find more salient than do whites. Note that this task differs from that of identifying the most salient issues for African Americans and Latinos. An issue might be very salient for one of these groups, but if it is also very salient for whites, its relative salience for the minority group is quite small.
54 | chapter three
Not surprisingly, prior studies have shown significant differences in the salience of issues to racial groups. African Americans place a higher priority on civil rights issues than do whites, and members of Congress are thus more responsive to the size of their African American constituency on these issues (Iyengar 1990; Hutchings 1998). In addition, the size of African American constituencies appears to be related to congressional voting on social welfare issues, which suggests that these issues may also be more salient for African Americans than for whites (Hutchings, McClerking, and Charles 2004; Whitby and Krause 2001). Comparisons of the issue priorities of whites and Latinos have identified three issue areas that are more salient for Latinos: education, economic security, and crime (Martin 2000; Uhlaner and Garcia 2002). It is not hard to imagine why these may be more salient issues for Latinos, given the economic, residential, and educational status of this group. Notably, we have already observed (fig. 3.5) that whites and Latinos differ substantially in their attitudes on spending to improve education, to reduce income differences, and to deal with crime. Whites and Latinos not only have quite different preferences on these issues, but the salience of these issues is relatively greater for Latinos. We can elucidate these differences more fully. The most common technique scholars have used to identify individuals’ issue priorities within a survey context is to ask respondents to name “the most important problem [hereinafter MIP] facing this country” (e.g. Smith 1985a and 1985b; Uhlaner and Garcia 2002). We can also refer to issue-specific measures of issue salience in which respondents are asked to judge the urgency of a single issue without comparing it to others. We first compared the open-ended responses of whites, African Americans, and Latinos to the MIP item used by the NES from 1974 to 2000. To do so, we calculated the proportion of each group’s respondents identifying welfare (or one of its synonyms), education, health care, crime, national defense, or the environment as the nation’s most important problem.17 We then divided the percentage of African American responses by the percentage of white responses in each policy category, and the percentage of Latino responses by the percentage of white responses in each policy category. Finally, we calculated the natural log of these ratios. If an equal proportion 17. We use the following response codes for NES variable number VCF0875b to form the binary indicators: welfare (60, 61, 63, and 90); national defense (700); the environment (150, 151, 153); health care (40); crime (340, 360), and education (20).
differences in political preferences and priorities | 55
figure 3 .6 . Relative issue salience across groups (American National Election Studies, 1974–2000).
of whites and African Americans identified an issue as the nation’s most important, this measure equals 0. If twice as many African Americans as whites identified an issue as most important, this measure equals 0.69, if half as many –0.69, and so on. The results are summarized in figure 3.6. Welfare policy has been the nation’s most important problem for more than twice as many African Americans as whites, and nearly three times as many Latinos as whites. Latinos were 2.5 times as likely as whites to identify crime as the nation’s most important problem, while African Americans were 1.5 times more likely. Latinos were about 2.5 times as likely as whites to mention education as the nation’s most important problem, while African Americans and whites were about equally likely to do so. Latinos were somewhat more likely than whites to name health care as the nation’s most important problem, while African Americans were a little less likely than whites to do so. However, other surveys find that African Americans are more concerned about health care than are whites. Finally, both national defense and the environment were much less often identified as national problems by African Americans than by whites, while the salience of these issues for Latinos and whites was fairly similar. In the case of African Americans, these results uncover one clear issue area that is more salient for African Americans—welfare—and two others
56 | chapter three
where we will see that the evidence is more mixed—health care and crime. These findings generally match the observations of political commentators who focus on the African American community.18 In the case of Latinos, these findings also largely confirm prior studies: education, economic security, and crime are more salient issues for Latinos than for whites. In addition, Latinos appear to be more concerned than whites about health care. These findings also largely reflect the state of Latinos’ priorities today. According to a 2004 survey of Latinos commissioned by the National Council of La Raza (NCLR) and conducted by Zogby International, 34% of Latinos thought that education was the most important issue facing the Latino community, while 22% identified the economy and jobs as the most important issue. Health care was also one of the top five issues identified. These findings largely match another measure of issue salience. The NAES asks to what extent respondents believe that poverty, the “number of criminals who are not punished enough,” and “the number of Americans without health insurance” are extremely serious problems. As figure 3.7 shows, only 29% of whites considered poverty an extremely serious problem, compared to 36% of Latinos and 46% of African Americans, a difference of 7 and 17 percentage points, respectively. A larger proportion of Latinos than whites thought that crime was an extremely serious problem, while (in some contrast to what we observed in fig. 3.6) about equal proportions of whites and African Americans viewed crime as an extremely serious problem. Finally, a far larger proportion of African Americans (55%) than whites (36%) thought that the number of individuals without health insurance was an extremely serious problem. In the NAES, a greater proportion of Latinos than whites were concerned about health insurance. Latinos’ focus on health care costs is also reflected in the NCLR survey, with 72% of those surveyed indicating that obtaining health insurance was a “big problem” for Hispanics in their area. Ultimately, we focus on three issues as being at least potentially more salient for African Americans than whites: welfare, crime, and health care. Likewise, we consider these three issues, along with education, as distinctively salient to Latinos. Next, we develop a measure of the overall importance of issues that incorporates both the relative importance of issues to each group and the ab-
18. According to Ronald Walters, in February 2004 five issues were most important for African Americans: jobs and economic well-being, including welfare reform; criminal justice; education; health care; and the war in Iraq. See his comments on the MSNBC Web site: http://msnbc.msn.com/id/4137295/.
differences in political preferences and priorities | 57
figure 3 .7 . Group salience of poverty, crime, and health insurance (2000 National Annenberg Election Survey).
solute importance of issues for each group. Our previous focus on relative importance was intended to establish a set of expectations regarding which issues might properly be evaluated under the pluralist standard of equality, which stresses that minority groups should enjoy more than proportional representation on issues that are more salient for them than they are for whites. Because we should not simply ignore the absolute importance of issues for groups, we developed an “impact score” that reflects both the relative and absolute importance of issues to groups. This “impact score” takes the percentage of all NES respondents in a minority group who identified an issue as most important and multiplies it by the ratio of this percentage to the percentage of whites who did the same (see fig. 3.8). This score shows that welfare and crime have the largest overall impact on African Americans, while welfare, health care, education, and crime have the greatest impact on Latinos in an overall sense. These impacts mirror what we observe in figures 3.6 and 3.7, but provide an important robustness check.
conclusion In summary, African Americans and whites differ over what government should do. They hold different ideological orientations and disagree about
58 | chapter three
figure 3 .8 . Impact scores for group salience (American National Election Studies, 1974–2000).
a host of racially related policies, including welfare, education, health care, and the environment. We also have established that Latinos and whites hold different ideological orientations and disagree about many policies, including welfare, education, health care, and crime expenditures. Moreover, we have seen that welfare, crime, and health care may be more salient to African Americans than whites, while education, income security, health care, and crime are more salient issues for Latinos than for whites. Finally, these differences in political preferences are not simply a reflection of differences in income. In fact, differences in political preferences are usually just as large even when we compare individuals within groups who have similar incomes. As we have observed time and again, Latinos, African Americans, and whites view the political world very differently. By and large, these groups want government to do different things and to focus its energies on different issues. These differences in political preferences create the possibility that government will follow some groups’ agendas and ignore others. When one group prefers more spending in an area and another prefers less, which group’s views carry the day? Testing whether the potential for political inequality between groups is realized in the policies of government is the task of the next chapter.
fo u r
Differences in Policy Representation In 2003, the poverty rate for whites was 8%, while the rate for African Americans and Latinos was approximately three times higher: 24% and 23%, respectively. Approximately 10% of white children lived in poverty, compared to 33% of African American children and 40% of Latino children. In the same year, 11% of whites did not have medical insurance. The proportion of African Americans without health insurance was nearly double that of whites (20%),1 and the proportion of uninsured Latinos was about triple that of whites (33%).2 According to the National Assessment of Educational Progress (NAEP) in 2000, 92% of whites ages 18–24 had completed high school, while only 84% of African Americans and 64% of Latinos had done so. Seventeen-year-old African Americans’ scores on the NAEP reading scale were 10% lower than whites’ scores, while Latinos’ scores were 9% lower. On the NAEP math scale, African Americans scored 10% lower and Latinos 7% lower than whites.3 From economic security to health care and education, racial and ethnic minorities in the United States fare far worse than whites. These disparities have many sources, but in its policy choices government can either attempt to ameliorate or blindly exacerbate them. Each year the federal government passes a budget and decides how much effort (if any) to commit to reducing 1. In December 2004, the American Journal of Public Health reported that 886,000 more African Americans died between 1991 and 2000 than would have died had equal health care been available. 2. Income and health insurance data are reported by the U.S. Census. See DeNavasWalt, Proctor, and Mills 2004. For the U.S. Census Bureau data, see the bureau’s Web site: http://www.census.gov/hhes/www/income.html. 3. See “Status and Trends in the Education of Blacks” on the Web site of the National Center for Education Statistics: http://nces.ed.gov/pubs2003/2003034.pdf.
60 | chapter four
income inequality through its income tax and entitlement programs, how much effort (if any) to devote to producing race-neutral health outcomes by subsidizing health insurance, and whether to direct additional resources toward minority educational programs. If it were up to members of minority groups, many government policies would be different. For instance, each year the Congressional Black Caucus releases an “Alternative Federal Budget” that it believes would better reflect the issue concerns and priorities of the African American community.4 Moreover, as we saw in the last chapter, African American and Latino citizens’ preferences for government spending often differ dramatically from whites’ preferences. In this chapter, we ask which groups “govern” the ultimate output of government actions—public policies. This chapter is devoted to answering three questions. First, do whites’ preferences predict federal policies better than the preferences of minorities? Second, how much better a predictor are whites’ preferences? Third, if whites are more often policy “winners,” is this simply a function of whites’ generally higher income levels and levels of political involvement?
“collective” and “dyadic” representation We begin our investigation of the extent of each group’s representation by looking at one specific type of representation, “collective” representation, which considers “representation in terms of institutions collectively representing a people” (Weissberg 1978, 535; emphasis in original). That is, we examine the degree to which the outputs of the policymaking process as a whole reflect citizens’ preferences. In later chapters, we explore the link between individual senators and members of the House of Representatives and their particular constituents. This relationship between individual members and their constituents, commonly referred to as “dyadic” representation, tends to be the focus of studies of policy representation. However, collective representation is an important place to start for a number of reasons. First, studying the connections between minority preferences and actual government policy is vital because the link between public opinion and the final decisions of government holds a fundamental place in democratic theory. As V.O. Key stated, “Unless mass views have some place in the shaping of policy, all the talk about democracy is nonsense” (Key 1961, 7). 4. For an example, see the Web site of the U.S. House: http://www.house.gov/scott/ legislative/issues/CBC%20ALT%20BUDGET%20FY2007.html.
differences in policy representation | 61
Public policies are at the heart of whether government ultimately responds to the preferences of citizens or does not. Policies, much more than the rollcall votes or other activities of members of Congress (MCs), affect the wellbeing of citizens—their education, health care, housing, and economic security, among other things. According to Verba (2003, 666), “Political equality in its fullest sense would be equal policy output.” Second, as Robert Weissberg notes, “whether or not a particular legislator follows his or her constituency is an important question, but this question is not necessarily the most appropriate one if we ask, ‘do representatives represent?’” (1978, 547). There may be important differences in the extent to which constituents are represented in the collective and dyadic senses. On one hand, some constituents may be better represented by their particular members of Congress than they are by the House of Representatives or the Senate as a whole. Liberals in Massachusetts or conservatives in Arizona may find their policy preferences perfectly represented by the liberal Senator Edward Kennedy (D-MA) or the conservative Senator John Kyl (R-AZ). However, the decisions made by the Senate as a whole and ultimately the policies forged collectively by the Senate, House, and president, may be far from what these constituents prefer. On the other hand, Weissberg (1978) shows that constituents are generally better represented by Congress as a whole than they are by their individual members. This is because it is difficult for any individual to represent almost seven hundred thousand constituents in a congressional district even if he or she wanted to. However, it is likely that of the 435 members of the House, one or more of them hold views and engage in activities that match relatively closely those of each of the seven hundred thousand constituents in a particular district. This may be especially true for African American or Latino MCs, some of whom explicitly seek to represent members of their group who are not in their home district. For example, Tate (2003a, 126) notes that “when Adam Clayton Powell of New York was one of two Blacks serving in Washington, he dealt with problems from Blacks all over the country, and not just those who had elected him, his Harlem constituents.” She also argues that “the Congressional Black Caucus would declare its mission as national” as opposed to the local concerns of each particular member’s district (Tate 2003a, 126–27). In the end, then, evidence that individual representatives act more in line with the preferences of their white constituents than their minority constituents may not mean that minority constituents are less well off in terms of collective policy outputs. A liberal Latino constituent represented by a
62 | chapter four
conservative MC is poorly represented in dyadic terms, but this does not guarantee poor representation in collective terms. As we noted in chapter 2, there are plenty of reasons to expect that whites are advantaged in the policy process, including their numerical superiority, greater resources, and greater involvement in public affairs. Our aim here is to test whether whites actually do find their preferences for policy realized more often than do African Americans or Latinos. However much we might anticipate that whites are advantaged in the policy process, this does not obviate the need for us to document the extent of whites’ advantage. Using measures of the public’s spending preferences from 1972 to 2002 and actual federal budget outlays from fiscal years 1972–2003, we evaluate whether whites are more likely than African Americans and Latinos to have their preferences realized in policy outcomes.
demand for government spending and governmental response In our first investigation of relative political representation, we examine whether African Americans’ and Latinos’ preferences for change or stability in policy are translated into actual policy change or maintenance to the same degree that whites’ preferences are. Examining the extent of policy responsiveness requires clear indicators of government policy and clear measures of public preferences about policy. These are not available for all government policies. Survey organizations do not typically ask the public its preferences on many of the technical details of government policies. Furthermore, policies often do not lend themselves to easy mapping onto survey items. For instance, it is difficult to say how much “more” gun control the Brady Bill established, and how much “more” environmental regulation the Clean Air Act produced. Federal spending decisions meet the needs of this type of analysis. Since 1972, survey organizations have regularly asked rather specific questions about the public’s desire for more, less, or the same amount of federal spending in various policy domains each year. These are specific questions about the preferred course of policy. In addition, the federal government must make decisions about how much to spend in various domains every year via the budget process. Thus, we can analyze specific public opinion questions and policy decisions at regular intervals. Furthermore, because both the survey questions and spending decisions pertain to specific issue domains, we can examine the link between preferences and policy across several issues.
differences in policy representation | 63
Our specific measure of policy outputs is annual federal expenditures as recorded in the U.S. budget (fiscal years 1972–2003) in billions of dollars.5 In this chapter, we analyze outlays in six policy domains: national defense, the environment, education, foreign aid, aid to big cities, and the space program. These policy areas were selected primarily because citizens have regularly been surveyed about their spending preferences in these domains. These are also domains that are not more salient for African Americans than whites (as we saw in chapter 3), and we wish first to compare racial groups’ representation in domains that are either equally salient for the groups or more salient for whites than minorities.6 We compare the policy representation of whites and African Americans on welfare, health care, and crime policy—the domains in which their preferences differ and where the issue appears more salient for African Americans—in chapter 6. For the same reason, we do not compare whites’ and Latinos’ representation on education, crime, health care, and welfare policy in this chapter.7 Our measures of public spending preferences are drawn from surveys the GSS and the Roper Organization conducted, which we introduced and discussed in chapter 3. Our data for African Americans’ preferences are primarily drawn from the GSS (1972–1994, 1996, 1998, 2000, and 2002), supplemented with three years of data from the Roper survey in years in which the GSS was not conducted (1979, 1981, and 1992) and before the Roper series was discontinued in 1994. In contrast, all our data for comparing the 5. See Office of Management and Budget 2004. See also the White House Web site: http://www.whitehouse.gov/omb/budget/fy2005/pdf/hist.pdf. 6. We did not discuss preferences for spending on foreign aid, space exploration, or big cities in chapter 3 because these were not queried in the NAES. In the GSS, the three groups held different spending preferences in these areas. There were no differences across the groups in terms of the salience of these issues. In fact, in each case less than 0.1% of NES respondents identified one of these issues as most important between 1974 and 1986, and no one mentioned any of these issues after that. These issues are not differentially salient for racial and ethnic groups, nor are they salient for any of the groups in an absolute sense. 7. To determine if budget line items should be assigned to an issue domain, we closely followed Wlezien (2004). Education spending included spending on elementary, secondary, and vocational education, higher education, and research and general educational aids; foreign aid included international development and humanitarian assistance, as well as international security assistance; aid to cities included community- development spending and area and regional development; welfare spending included housing assistance, food and nutrition assistance, and other income security. The remaining domains included all spending itemized in the budget function/subfunction.
64 | chapter four
representation of whites and Latinos is drawn from the Roper series, because the GSS did not include a specific survey question to identify Latino respondents until 2000. These measures of Latino opinion begin in 1979, after the Roper series began to query respondents about whether they were “Hispanic,” and conclude in 1994.8 By mapping the opinion data onto the expenditure data, we identify whether individuals were on the “winning” side of federal budget decisions in a series of policy domains in the year the individual was surveyed.9 Specifically, if a respondent desired more (less) spending in an issue area and federal outlays increased (decreased) more than 5% in constant dollars, then we call that respondent a “winner” in that issue area. For that same issue and year, we call respondents who desired no change in spending “losers.” Finally, we affectionately call respondents who preferred less (more) spending “big losers.”10 We consider citizens who desired no increase in spending in a policy domain winners if the percentage change in outlays in the following fiscal year was between a 5% increase or decrease.11
whites, african americans, and “policy winning” We begin by assessing whether African Americans are less likely than whites to be on the “winning side” of policy decisions. We simply tabulate the proportion of whites and African Americans who are winners, losers, and big losers in each policy domain. The results of these tabulations indicate that in some issue areas African Americans are less likely than whites to be policy winners (see fig. 4.1). In three of the six policy domains—defense, aid to big cities, and space exploration—African Americans are clearly less likely to be winners than whites. Moreover, African Americans were much more likely than whites to be big losers on spending for defense, aid to big 8. We opted to pool the smaller GSS and Roper survey samples over time rather than using the larger NAES sample because the pooling approach allows us to gain a sense of whether racial minorities have generally been policy winners or losers in a domain averaged over a number of years, whereas the NAES only permits us to form comparisons between racial groups for fiscal year 2001, for which the budget may or may not be indicative of spending decisions in prior and subsequent years. 9. For a similar approach see Hajnal, Gerber, and Louch (2002). 10. One difficulty in using this approach arises when spending does not change. In those years, the measure of policy “winning” does not have the same range, because there are no “big losers.” 11. We experimented with various cutoff points for defining policy winners and losers. The 5% cutoff we use here tends to generate more conservative results.
differences in policy representation | 65
figure 4 .1 . Percentage of policy winners and big losers, whites and African Americans, 1972–2002 (Office of Management and Budget 2004; General Social Surveys, 1972–1994, 1996, 1998, 2002; Roper polls, 1979, 1981, 1992).
cities, and space exploration. In one area, the environment, whites have a narrower advantage. In two areas, education and foreign aid, African Americans are a little more likely than whites to be winners. Whites were quite a bit more likely to be winners in outlays for defense— nearly 40% of whites were winners in this domain, compared to around 31% of African Americans. A sizeable proportion (18%) of African Americans were “big losers” on defense spending, compared to 14% of whites. In general, these individuals preferred less spending on defense when defense spending increased significantly. On spending for big cities and space exploration, white winners outpaced African American winners somewhat less. Where 38% of whites were winners on aid to big cities, 32% of African Americans were winners. Moreover, a greater proportion of African Americans (15%) than whites (11%) were big losers in this domain. Most dramatically, about one in three whites were winners on space exploration, but only about 16% of African Americans were winners. Moreover, one in three African Americans were “big losers”—most of whom wanted less spending when outlays rose over 5%. Not only were “big losers” double the number of winners for African Americans, but the percentage of African American “big losers” was
66 | chapter four
14 points greater than the percentage of white “big losers.” Of the domains we examined, African Americans’ preferences are reflected least well in the domain of space exploration, both in absolute terms (the raw percentage of African American winners compared to other domains) and in relative terms (the difference in African American winners compared to white winners). On the environment, whites hold a smaller advantage over African Americans—46% of whites were winners, while about 44% of African Americans were winners. In chapter 5, we probe this small difference further, asking whether whites and African Americans are equally proximate to their representatives using votes identified by the League of Conservation Voters as key roll calls. As we will see, whites’ preferences tend to be much better reflected in Congressional roll-call voting on the environment. Finally, there are two domains where African Americans are somewhat more likely to be policy winners: spending on education and foreign aid. Three factors may account for African Americans’ unexpected advantage regarding education. First, these results may suggest that it is a more important domain for this group than our figures in chapter 3 reflect. For instance, a recent book-length study of African American and Latino representation in four northeastern cities focused on these groups’ representation in two policy domains: education and public safety (Burns 2006). Its justification for doing so was that “these issue areas strongly affect traditionally excluded groups” (8). Moreover, as noted above, political scientist Ronald Walters recently identified education as one of the five most important issues for the African American community.12 Second, it may be that office holders respond to the priorities of minority groups in general, rather than group by group. Since many of the issues Latinos and African Americans care about more than whites do are the same (e.g., welfare, crime, health care), office holders may assume that any issue especially salient to Latinos is also especially salient to African Americans. As we saw in chapter 3, education is more often salient to Latinos compared to whites, but not to African Americans, at least according to our data. Officeholders may be assuming that education is also especially salient to African Americans. Note that the first two factors that may account for this unexpected finding suggest that education is either actually more salient for African Americans than whites (even though our data do not show this) or that officeholders perceive it to be distinctively salient to African Americans. A 12. See, http://msnbc.msn.com/id/4137295/.
differences in policy representation | 67
third account accepts that education is not more salient to African Americans compared to whites and that officeholders realize this. However, African Americans may benefit from Latinos’ more intense preferences on education issues. We will see in chapter 6 that on issues distinctively salient for minority groups, those groups tend to be equally represented—or even better represented—compared to whites. In fact, we will see that Latinos tend to win almost as much as whites do on education spending. African Americans may benefit from this because African Americans’ preferences for spending on education are far closer to those of Latinos than to those of whites. Therefore, if Latinos are winners on education spending, many African Americans will also be winners. Finally, a slightly larger percentage of African Americans than whites were winners on foreign aid: 25% to 22%. However, neither whites nor African Americans do particularly well when it comes to foreign aid. We found fewer winners on foreign aid than on any other issue we examined. In fact, about 40% of each group were big losers on foreign aid. So, the one domain in which African Americans are somewhat more likely than whites to be winners is also the domain in which African Americans are least likely to be winners at all. For the most part, both African Americans and whites wanted less or the same amount of spending when outlays remained the same or rose significantly. This pattern is presumably a product of Americans’ tendency to vastly overestimate the amount that the United States spends on foreign aid each year (Gilens 2001).
whites, latinos, and “policy winning” Next, we assess whether Latinos are less likely than whites to be winners with regard to policy decisions. There are reasons to believe that Latinos may fare better than African Americans relative to whites, and other reasons to believe that they may fare worse. On the one hand, we saw in chapter 3 that, compared to the preferences of African Americans, Latinos’ preferences generally are less different from those of whites. This relative similarity dampens differences in the frequency of winning between whites and Latinos. On the other hand, Latinos are confronted by unique obstacles to political representation that African Americans do not face. The most important of these is that many Latinos are not citizens and thus cannot legally vote. In addition, English is not the first language for many Latinos. These factors limit Latinos’ electoral clout and their likelihood of communicating their preferences to elected officials, which are two important routes to political representation.
68 | chapter four
Our approach to comparing whites’ and Latinos’ rates of winning parallels that which we have just seen for whites and African Americans, with two departures. First, we rely exclusively on the Roper data for these analyses because the GSS did not clearly identify Latino respondents for most of the period under study. Second, in addition to reserving the analysis of welfare, health care, and crime spending, we reserve discussion of spending on education for chapter 6 because this issue is more salient for Latinos than for whites. Respondents are coded as Latino if they identified themselves as “Hispanic” in the Roper surveys, and they are coded as white if they were self-identified whites who did not also identify themselves as Hispanic. Because the sample sizes of Latinos are smaller than those of African Americans in our comparisons of whites’ and African Americans’ rates of winning, we are less confident in these results. Despite the need for caution, the results prove enlightening. When we tabulate policy “winners” among whites and Latinos, we again find that in some domains Latinos are less likely to be policy winners (see fig. 4.2). Latinos are quite a bit less likely than whites to see their preferences realized in policy on national defense and environmental regulation, and somewhat less likely to be policy winners on foreign aid. On the other hand, Latinos are roughly equally likely to be policy winners on aid to big cities and space exploration.13 Looking at the specific results, we see some of the patterns we saw when comparing African Americans and whites. In the three domains where a greater percentage of whites than Latinos were winners, the gap was between 3 and 7 points. White winners outpaced Latino winners by just under 7 points on defense spending (36.3% to 29.7%), by 5 points on environmental spending (45% to 40%), and by 3 points on foreign aid (34% to 31%). These differences are generally smaller than the differences between whites and African Americans that we observed in figure 4.1. Also in contrast to African Americans, the percentage of Latino “big losers” substantially exceeds the percentage of white “big losers” in just one domain, aid to big cities. In summary, Latinos were less likely than whites to be winners in three of
13. The predicted probabilities for whites are a bit different than those in figure 4.1. This is likely due to different data sources (GSS versus Roper) and different time periods of study. For our primary purpose of assessing the relative representation of different groups, the relative probability of winning between whites and Latinos (or whites and African Americans), rather than the absolute probability of winning, is the critical measure.
differences in policy representation | 69
figure 4 .2 . Percentage of policy winners and big losers, whites and Latinos, 1979–1994 (Office of Management and Budget 2004; Roper polls, 1979–1994).
five domains and were more likely to be big losers in a fourth domain (aid to big cities).
income and policy winning Returning now to the comparison of whites and African Americans, recall that in the four domains in which whites win more often, the percentage of white winners outpaced the percentage of African American winners by 2 to 10 points. Each of these differences (except environmental spending) is statistically significant at the .05 level. Substantively, does this 2–10-point gap constitute a “big” difference, one we should be concerned about? One way to judge this is to compare the difference in the probability of winning for two demographic groups where we would anticipate that one group should be much more politically influential than the other. There has long been a belief and concern that groups with higher incomes exert greater influence in politics than those with lower incomes (e.g., Schattschneider 1960; Gilens 2005; Bartels 2008). In order to compare the racial differences in policy representation that we have found to income differences in representation, in
70 | chapter four
figure 4 .3 . Percentage of policy winners among high- and low-income respondents (Office of Management and Budget 2004; General Social Surveys, 1972–1994, 1996, 1998, 2002).
the same four domains we compared predicted probabilities of being policy winners between GSS respondents whose income in constant dollars was above the seventy-fifth percentile of the income distribution and respondents who were located below the twenty-fifth percentile of the income distribution (see fig. 4.3).14 Although the higher income group (white checked columns) was sometimes more successful in getting its preferred outcome, this was not always the case. For outlays on the environment and big cities, higher-income respondents were not systematically more likely to be policy winners at all. While higher-income respondents were more likely to be policy winners in two domains—defense and space exploration—in these domains the racial differences in the likelihood of winning we have uncovered were equal to or greater than the income differences. In general, the “policy winning gap” between African Americans and whites is greater than that between the wealthy and the poor. 14. We employ only the GSS data because the Roper data report income using a different scale, one that is not comparable over time. We do not report big losers for presentational reasons, but note that lower-income earners were somewhat more likely to be big losers on defense, crime, and space exploration spending.
differences in policy representation | 71
Our analysis of race and policy representation thus far only indicates general differences between African Americans and whites, without taking income into account. We know that African Americans tend to earn lower incomes than whites, and, as we have just seen, lower-income earners tend to be less well represented in some domains (see also Gilens 2005; Bartels 2008). To get a sense of the direct effect of race on an individual’s probability of being a policy winner, in the four domains where we have observed that whites are more likely than African Americans to be policy winners, we tabulated the proportion of each racial group that was a policy winner among respondents with high (greater than the 75th percentile), middle (less than the 75th percentile and greater than the 25th percentile), and low (less than the 25th percentile) incomes, where incomes were measured in constant dollars. These tabulations indicate the extent to which race and related factors such as relative voting power (Bartels 1998) are directly related to the probability of being a policy winner, over and above the effect of race on income. The results of these tabulations suggest that both income and race play roles in winning and losing (see fig. 4.4). All four of the policy domains include evidence that whites were more likely to be winners even within income groups. For example, figure 4.1 showed that 40% of whites were winners on spending for defense, but only 31% of African Americans were winners. This difference remained about the same when controlling for income. Looking only at high-income earners, the gap is the same: 41% of whites to 32% of African Americans (see fig. 4.4a). The same is basically true among middle-income earners, as 38% of whites were winners, compared to 30% of African Americans (see fig. 4.4b). Among low-income earners, the gap decreased, but even here a gap remained, with 35% of whites winners, compared to 31% of African Americans (see fig. 4.4b). Other policy domains exhibit a similar pattern. Although income is related to policy winning, race continues to be powerfully linked to winning and losing. African Americans’ incomes help little at all to explain why members of this group are less likely to be policy winners.15
are policy winners just more attentive citizens? Our evidence that whites win more often than minorities may simply be spurious in a way we have ignored thus far. Government policy decisions 15. We did not conduct a similar analysis of Latinos and whites because the Roper polls we use to analyze whites and Latinos do not measure income in constant dollars.
figure 4 .4 . Percentage of policy winners and big losers, by income. (Top) High-income whites and African Americans; (middle) middle-income whites and African Americans; (bottom) low-income whites and African Americans. (Office of Management and Budget 2004; General Social Surveys, 1972–1994, 1996, 1998, 2002).
differences in policy representation | 73
may be directly responsive to developments that expose policy shortcomings, and citizen opinion may be responsive to these events in similar ways, creating a spurious correspondence between citizen opinion and policy change. For example, after a major environmental disaster, such as the Exxon Valdez oil spill, public preferences for spending on the environment may increase. This disaster may also convince public officials to increase spending regardless of any change they observe in public preferences. Moreover, if whites are more likely than African Americans and Latinos to be interested in and pay attention to such events, this potentially spurious relationship between opinion and policy will be greater for whites. If this occurs, the disparities in winning we have uncovered may be due to a combination of the effect of actual events on policymaking and differences among respondents in attention to the real world. Ideally, we would like to incorporate citizens’ level of political attention or knowledge to account for this. Unfortunately, neither the GSS nor the Roper polls regularly queried their respondents about their interest in and knowledge of political affairs or world events. To test for spuriousness in our findings that whites are more likely to be policy winners for spending on defense, aid to big cities, the environment, and space exploration, we repeated our tabulations in figure 4.1 for the subset of white and African American respondents who reported not voting in the most recent presidential election. The logic here is that nonvoters tend to be less interested in and attentive to politics, and so presumably less likely to revise their spending preferences on the basis of world events (Bennett and Resnick 1990; Verba, Schlozman, and Brady 1995). If we find that even among the subset of GSS respondents who are nonvoters whites are more likely to be policy winners, we will gain confidence that racial differences in attention to public affairs are not driving our results. Unfortunately, this approach excludes all of the Roper respondents because those surveys did not query about respondents’ turnout. The unhappy consequence of this is we cannot perform this robustness check for Latinos because we rely exclusively on the Roper data for those analyses. As figure 4.5 shows, even among less politically interested citizens (i.e., nonvoters), whites remain more likely to be policy winners compared to African Americans. In all four issue domains in which whites were advantaged, African American nonvoters are systematically less likely to be policy winners compared to white nonvoters. Moreover, comparing these results to those in figure 4.1 reveals that the magnitude of the difference between nonvoting whites and African Americans in being policy winners appears generally to rival that between all whites and African Americans.
74 | chapter four
figure 4 .5 . Percentage of policy winners and big losers, white and African American nonvoters (Office of Management and Budget 2004; General Social Surveys, 1972–1994, 1996, 1998, 2002).
Finally, in unreported tests controlling for related factors such as education, income, age, gender, and ideological orientation, African American nonvoters remain less likely than white nonvoters to be policy winners in the same domains.
implications To this point, we have demonstrated that African Americans and Latinos are often less likely than whites to be policy winners. Because these comparisons have not taken into account the racial and ethnic composition of the national population, they are best characterized as evaluations of raceconscious egalitarianism—testing whether racial and ethnic groups are equally represented as groups. According to a strict application of this standard of equality, African Americans and Latinos appear somewhat unequal. The average white American is, more often than not, more likely to see his or her preferences realized in government action than is the average African American or Latino. Our analysis of differences in rates of policy winning adds to existing research in two important ways. First, studies comparing the representation
differences in policy representation | 75
of different groups in policy have tended to study different income groups (Gilens 2005). We show the importance of supplementing studies of income differences in policy representation with studies of racial and ethnic differences in representation. Both race/ethnicity and income are related to political representation. Studies of political representation should take into account the racial, ethnic, and income differences in representation that we have demonstrated. Second, our finding that African Americans, and (somewhat less so) Latinos generally win less than do whites also has important political implications. Since the legislative branches are elected bodies and whites far outnumber minorities, we might anticipate that elected bodies will produce outcomes that favor whites. However, other institutions in the policymaking process may “correct” such unequal representation. For instance, Frymer (2003, 483) has documented the intervention of the judicial branch to protect the interests of racial minorities “when elected officials won’t.” Our findings indicate that in these domains any disparities in representation in the legislative branches generally are not offset by the other bodies of government. While this may happen in some areas, our analyses in this chapter show that not all disparities in representation created by the electoral system are corrected in the policy process. Our approach in this chapter provides an important view of the relative representation of racial and ethnic groups. However, this is only one view. The picture we see from the perspective of collective rather than dyadic representation does not show us everything about the relative representation of these groups. Our search for clear indicators of policy preferences, policy outputs, and a close connection between them has led us to focus on preferences for federal spending and subsequent federal outlays. The analytical advantages of this approach necessarily limit our investigation by excluding issues that do not clearly relate to spending. Important and controversial issues like abortion, civil rights, gun control, immigration, and same-sex marriage have unclear budgetary implications. It is unclear whether a citizen who prefers that same-sex marriages be legalized would prefer more, less, or the same amount of spending related to same-sex marriages, for example. In addition, our measures of preferences only allow citizens to indicate whether they want more or less spending (or the same amount). They do not allow them to express how much more or less. Therefore, our ability to assess the connection between preferences and government response is limited. For example, if an African American citizen wants spending on big cities to double and spending increases by only 6 percent, our method
76 | chapter four
considers this citizen a winner. However, this citizen’s preference is clearly not perfectly met. In a related way, even if a citizen wanted more spending on education, and federal outlays on education actually did increase, we do not know if the citizen is happy with the way the government spent the money. A citizen who wanted more bilingual education programs probably would be unhappy if the government increased funding for basic research on physics at universities. Finally, since the government as a whole reaches a single decision that affects everyone, it is impossible to examine how representation might differ under varying circumstances that we know will affect representation, like descriptive representation or the size of minority populations, which vary across districts but not over the nation at any given time. Although our approach here does not tell us everything, the approaches taken in the next chapters help to fill out our picture of relative representation of African Americans, Latinos, and whites by focusing on the blind spots of this chapter’s approach. This chapter takes the first step in our investigation. As we have seen, there are significant disparities in policy representation regarding government budget decisions. Whites appear to enjoy better representation than African Americans and Latinos regarding policy in many of the issue domains that are not more salient for minorities. In domains where there are differences in the frequency of winning, these differences are often much larger than the difference in winning rates between wealthy and less wealthy citizens. Moreover, differences in policy winning do not appear to be an artifact of whites’ greater attentiveness to political affairs. In the next chapters we unpack this general finding to discover some of the sources of unequal representation. In chapter 5 we widen our scope by turning to dyadic representation, exploring the connections between individual legislators and their constituents. Unfortunately for minority groups, the story is unlikely to improve since “collective representation will never be worse than dyadic representation” (Weissberg 1978, 547). We begin in the next chapter by examining whether the three groups are unequally represented by their elected representatives’ behavior. We know that such a step is necessary because, as we have seen, government policy often better reflects whites’ preferences than those of African Americans and Latinos.
fi ve
Differences in Legislative Representation In the Federalist No. 52, James Madison argued that the proposed Congress, and in particular the House of Representatives, “should have an immediate dependence on, and an intimate sympathy with, the people” (Madison, Hamilton, and Jay 1987, 323–4). However, not all “the people” feel an “intimate sympathy” with their representatives. There are significant differences in racial and ethnic groups’ perceptions that elected officials are responsive to their group, as figure 5.1 shows. In the 2002 National Election Studies, conducted during the time period we analyze, 43% of the nation’s African Americans and 32% of the nation’s Latinos agreed that “[p]ublic officials don’t care much what people like me think,” while just 27% of whites held this view. In that same year, 39% of African Americans and 34% of Latinos agreed that “[p]eople like me don’t have any say in what the government does,” compared to just 24% of whites. Finally, 44% of African Americans and 39% of Latinos disapproved of Congress’s performance, compared to 36% of whites. In this chapter, we evaluate whether these perceptions reflect reality, examining the extent to which the nation’s legislative bodies are equally attentive to the preferences of all “the people.” In the last chapter, we focused on collective representation, the connection between citizens across the country and the policies the government as a whole produces. In this chapter, we explore the connection between individual members of Congress (MCs) and their particular constituents, or dyadic representation.1 This shift offers several benefits. First, it enables us
1. For other studies of dyadic representation, see Miller and Stokes 1963; Achen 1978; Erikson 1978; Ansolabehere, Snyder, and Stewart 2001; Bartels 1991 and 2008.
78 | chapter five
figure 5 .1 . Group differences in alienation (2002 American National Election Study).
to use methods that complement the previous chapter’s policy analysis by avoiding the limitations of that type of analysis. Second, studying representation within the nation’s legislatures permits us to assess one important mechanism by which minorities’ preferences become disadvantaged in the policy process: legislative roll-call voting. Chapter 4 described differences in policy representation but did not tell us much about how these disparities come about. The nation’s legislatures offer a natural first place to look for the origins of differences in policy representation. Members of Congress are elected under plurality electoral rules by districts that are often home to far more whites than African Americans and Latinos. Once in office, they make decisions by simple majority or, in the case of the Senate, modest supermajority rule. These are just two features of Congress that would seem to work to the disadvantage of minority groups. Third, by looking at both collective and dyadic representation, we explore different facets of representation, an inherently multifaceted phenomenon. If elections, the hallmark of democratic politics, help establish and maintain the connection between the governed and the government, we should see a connection between elected officials and those who elect them. We as citizens may have little influence over a member of Congress
differences in legislative representation | 79
from another state, but we can at least vote against our own member in the next election. Ultimately, if we find similar results by looking at different facets of representation and using complementary empirical methods, we will have more confidence in our findings and have a more comprehensive picture of relative representation among these groups. Finally, this approach allows us to explore a matter that is critical to our study: the ways in which variation in state or district conditions such as the racial or ethnic composition of states affects the decisions of federal officials. This enables us to evaluate the performance of the political system against an additional standard of political equality. Our analyses in the previous chapter compared the groups’ frequency of policy winning against a strict standard of race-conscious egalitarianism—that groups should have a relatively equal likelihood of being on the winning side of outcomes—but could go no further. Studying groups’ legislative representation permits us to assess the performance of the political system against a proportional standard of political equality. Moving from chapter 4’s national-level analysis to state- and congressional district-level analysis enables us to examine relative representation in a number of different settings, simply because some districts have very few African Americans or Latinos, while in other districts African Americans or Latinos make up a substantial proportion and sometimes even a majority of the constituency. Thus, we can see whether minorities are equally represented in legislators’ voting decisions when they comprise a larger share of an electoral district. We structure the present chapter around these two standards of equality, first comparing groups’ legislative representation against race-conscious egalitarianism, and then against a proportional standard of equality. We leave our discussion of the pluralist standard of political equality for the next chapter. Our assessment of the groups’ representation against each of these standards of equality adopts two measures of legislative representation— the responsiveness and proximity of legislators relative to racial and ethnic groups. Since representation is a complex phenomenon, scholars must examine different facets of representation. Following the lead of Miller and Stokes (1963), many works examine the extent to which representatives’ roll-call votes relate to the average opinion in their district (Achen 1978; Erikson, Wright, and McIver 1993; see also Erikson 1978; McCrone and Kuklinski 1979; Bullock and Brady 1983). Among these opinion/roll-call studies, one approach evaluates legislators’ responsiveness as a group to district opinion, estimating the extent to which liberal legislators represent liberal
80 | chapter five
constituencies and conservative legislators represent conservative constituencies (e.g., Achen 1978; Ansolabehere, Snyder, and Stewart 2001; Erikson and Wright 2000). Other studies examine the aggregate “distance” between individual legislators’ actions and all their constituents’ preferences, a concept Achen termed proximity (Powell 1982; Wright 1978). According to Achen (1978), these concepts of responsiveness and proximity capture different, important features of representation—whether, respectively, “the representative system adapts to citizen preferences,” and “everyone’s voice [is] counted equally” (490, 481). For reasons that will become clear, adopting these measures of representation naturally leads us to investigations of both national legislative institutions, the U.S. Senate and the U.S. House of Representatives. The Senate and the House each offer unique analytical advantages for assessing the extent of the three groups’ relative representation. Although many have studied representation in the Senate generally (e.g., Wright 1989; Erikson 1990; Wood and Hinton Andersson 1998; Fowler 2005), most studies of minority representation examine the House, meaning we know far less about differences in representation in the Senate. Yet the Senate may be especially unresponsive to minorities’ interests, since African Americans and Latinos typically constitute less than 20 percent of state populations and thus there are no majority-African American or Latino states where African Americans’ or Latinos’ interests may be especially well represented. Historically, there also have been very few African American or Latino senators. In contrast, analyzing the House of Representatives enables us to assess how larger concentrations of minority constituents and the race or ethnicity of MCs affect groups’ substantive representation (we analyze the impact of MCs’ race and ethnicity in chapter 7).
race-conscious egalitarianism To evaluate the groups’ legislative representation against the standard of race-conscious egalitarianism, we follow most studies by focusing on the relationship between citizens’ preferences and their elected representatives’ voting behavior.2 As we relate citizens’ preferences to their representatives’ decisions, if groups are politically equal by a strict, race-conscious 2. Of course, roll-call voting constitutes only one way for MCs to represent their constituents. For discussions of other representational activities MCs can engage in, see Canon (1999) and Tate (2003a).
differences in legislative representation | 81
egalitarian standard, then minorities’ preferences and whites’ preferences aggregated within electoral districts will be equally effective in predicting their representatives’ decisions. Second, if the groups are politically equal under this standard, then individual members of racial and ethnic groups will be on average equidistant from their representatives. At the very least, under this standard of equality, representation should be equal across groups when minorities constitute some substantial share of the population, making minorities’ representation more than proportional. Race-Conscious Egalitarianism and Responsiveness | The most common approach to studying representation assesses legislators’ responsiveness to constituent opinion. We start by following in this tradition, analyzing senators’ roll-call votes as a function of their African American, Latino, and white constituents’ preferences. This specific approach overcomes three of the limits of the previous chapter’s policy analysis. First, it allows us to explore issues we could not explore in that analysis (i.e., we can move beyond federal spending decisions). Second, we explore citizens’ general ideological perspectives, which enables us to take magnitude into account much more than does the policy approach. In the spending policy analysis, constituents could only express a preference for more, less, or about the same amount of spending, not how much more or less spending they would prefer. In the present approach, constituencies can be more than just “liberal” or “conservative.” They can be extremely liberal or moderately liberal. This approach also allows MCs to vary widely (and continuously) in the degree to which they are liberal or conservative. Third, the substance of the roll-call votes is much more clear than is true for simply “more” or “less” spending. For example, we analyze one roll-call vote related to boosting unemployment benefits. We compare this to whites’ and minorities’ preferences regarding federal attempts to reduce income differences between rich and poor Americans. Both the roll-call vote and the preference measure are more specific than the rather general preference for more/less/the same spending and the general budgetary response of more spending. Before we examine specific roll-call votes though, we begin by analyzing senators’ votes across a full range of issues. To measure overarching patterns in senators’ roll-call voting in the 107th Congress (2001–2002), we use a summary measure known as w-nominate coordinates. These coordinates are generally agreed to reflect a legislator’s mean roll-call behavior on a liberal-conservative social welfare spectrum ranging continuously
82 | chapter five
from –1 to +1, with larger values indicating more conservative roll-call voting patterns.3 The responsiveness approach requires district-level measures of citizens’ preferences. In forming these measures, prior studies of representation tend to treat citizens as an undifferentiated group; however this design can be adapted to study responsiveness to groups within electoral districts. For example, Bartels (2008) used a variant of this approach to determine that senators respond disproportionately to constituents with higher incomes. As a practical matter, though, the data requirements of the responsiveness approach often are difficult to meet when studying groups within geographic constituencies. Fortunately, the NAES provides relatively representative and reliable state-level measures of the three groups’ ideological orientations. As noted above, these data include 26 states with at least 50 African Americans and 21 states with at least 48 Latinos.4 In addition to the size of the sample (more than 57,000 respondents), which far exceeds that of most surveys, the NAES sampling method offers another advantage for creating state-level measures of opinion. The NAES used a Random Digit Dialing (RDD) procedure, rather than a stratified clustered sample (as used by the National Election Studies).5 The RDD design more closely approximates the ideal of making everyone in 3. w-nominate coordinates are generated by placing legislators and the midpoint between non-unanimous roll-call alternatives in a two-dimensional ideological space, making the assumption that legislators always vote for the nearest alternative, and then iteratively relocating legislators and roll-call midpoints until the number of rollcall “errors” is minimized (see Lewis and Poole 2004 for a more complete description). These coordinates correlate highly with alternative, summary measures of legislator roll-call behavior like interest group ratings and Heckman-Snyder scores (Burden, Caldeira, and Groseclose 2000). This version of the w-nominate scores is appropriate because it measures behavior in a single legislative term, independent of other terms served. Analyses of congressional roll-call voting have regularly used various versions of w-nominate scores as dependent variables (e.g., Ansolabehere, Snyder, and Stewart 2001; Bartels 2008). 4. The states we use in our comparisons of whites and African Americans include AL, AR, CA, CT, FL, GA, IL, IN, KY, LA, MA, MD, MI, MO, MS, NJ, NY, NC, OH, OK, PA, SC, TN, TX, VA, and WI. The states we use in our comparisons of whites and Latinos include AZ, CA, CO, CT, FL, GA, IL, MA, MI, MO, NC, NJ, NM, NV, NY, OH, OR, PA, TX, VA, and WA. 5. The stratified sampling method is typically designed to draw a sample that is representative of the nation, but not states (see Brace et al. 2002). For example, a survey may sample several respondents in a major population center like Detroit, but not in other areas in that state. Such a sample would not necessarily be representative of all of Michigan.
differences in legislative representation | 83
a state equally likely to be sampled, thereby generating more representative state-level samples. In fact, the NAES state samples match census data fairly closely, suggesting that these samples are quite representative.6 In addition to being representative, as we noted in chapter 3 (see note 6), under standard definitions of reliability state-level mean white ideological orientations were highly reliable; African Americans’ and Latinos’ state-level ideological orientations were less reliable. To compensate for the lower reliability of the measures for African Americans and Latinos, when we anticipate that African Americans and Latinos are likely to be underrepresented, we use an errors-in-variables estimator to account for measurement error. This boosts the estimated relationship between African Americans’ and Latinos’ opinions and senators’ decisions, thereby minimizing the amount of unequal responsiveness we will find. This “stacks the deck” in favor of equal representation across groups. If we continue to find inequality under these conditions, we will be all the more confident in our conclusions. To summarize, we calculated the mean opinions of whites and African Americans in 26 states, and the mean opinions of whites and Latinos in 21 states yielding African American Ideology, Latino Ideology, and two measures of White Ideology. The roll-call and opinion measures are coded with conservative alternatives higher, so positive estimates for white and African American/Latino opinion indicate positive responsiveness. To determine the responsiveness of senators’ votes to the preferences of African Americans, Latinos, and whites, we model each senator’s votes as a function of his or her mean white and mean African American or Latino constituents’ opinions. If senators’ votes respond more to whites’ preferences, the estimated relationship for whites’ preferences will be greater than that for African Americans’ or Latinos’ preferences. If senators are not responsive to a particular group’s preferences at all, the estimated parameter for that group will be zero. These models are not intended to explain senators’ votes, but to describe the correspondence between senators’ votes and constituents’ preferences. Other factors surely affect senators’ votes, 6. To evaluate the representativeness of the NAES state-level samples, we compared the demographics of the NAES data to the reported state demographics in the 2000 U.S. Census. We found that in the twenty-six states on which we focus our analyses of African Americans, the proportions of the population that is African American in the NAES and in the census correlate at .98. The state median income correlates with the (collapsed) NAES income self-placement scale at .85. The percentage of state residents who live in urban settings and have college degrees in the NAES and the census correlate at .88 and .87, respectively.
84 | chapter five
but our present task is simply to test whether the end product of those other factors, actual votes, better reflect their white or African American/Latino constituents’ preferences.7 We first estimate a simple model of w-nominate scores as a function of just white ideology or African American ideology, along with MCs’ party affiliations (see the left side of table 5.1). Comparing the first two columns in table 5.1, we observe that white ideology is a better predictor of senators’ w-nominate scores than is African American ideology (compare coefficients of 1.09 for whites and 0.67 for African Americans).8 This no longer appears to be true after we account for the lower reliability of our African American ideology measure using an errors-in-variables estimator in the third model (see column 3, where the African American ideology parameter is 1.28).9 However, in some states, especially in the South, both whites and African Americans are more conservative than their counterparts in other states, so white ideology is related to African American ideology.10 African American ideology might be spuriously related to senators’ roll-call voting based on its correlation with white ideology. To control for this, we included both white and African American ideology in a single model (see column 4). White ideology once again becomes a better predictor of senators’ voting behavior, relative to African American ideology.11 In fact, the estimate for
7. For an example of this general approach as applied to the representation of business interests, public opinion, and other factors on foreign policy, see Jacobs and Page 2005; see also Bartels 2008. 8. Out of concern that the two observations for each state exhibit correlation in the independent variables, we also estimated separate models of each state’s junior and senior senators. In these models, once again senators are responsive to variation in white but not African American ideology. 9. We also estimated this model using senators’ second-dimension w-nominate scores as the dependent variables. In some historical periods, these second-dimension scores appear to measure legislators’ voting behavior on racial issues (Poole and Rosenthal 1997). We found that neither white nor African American ideology predicted senators’ second-dimension scores in the 107th Senate. 10. The correlation coefficient between them is .50. 11. We also randomly sampled among whites to create measures of white and African American ideology that were based on the same state sample sizes. Estimating the model reported in table 5.1, column 3 using these measures yielded a coefficient for white ideology that was somewhat smaller and statistically significant (p < .01), while the coefficient for African American ideology remained insignificant. To guard against the possibility that white and African American ideology are collinear, we also regressed senators’ w-nominate scores on the difference between white and African American ideology, finding that as whites become more conservative relative
1.15** [.06] –4.11** [.69]
1.28** [.06] –4.50** [1.51]
1.28* [.50] —
.67* [.28] —
1.29** [.06] –2.67** [.84]
—
—
.53
1.16** [.06] –4.21** [.82]
1.06** [.26] .06 [.28] —
—
—
OLS
Errors in Vars —
(4)
(3)
1.15** [.07] –4.38 [1.40]
1.09* [.47] .09 [.83] —
.53
.97
Errors in Vars
(5)
Sources: 2000 National Annenberg Election Survey; Lewis and Poole 2004. Notes: Dependent variable: Senators’ w-nominate scores. * p < .05; ** p < .01; standard errors in brackets. “Errors in Vars” denotes errors-in-variables regression.
Constant
Republican MC
—
1.09* [.22] —
—
—
White ideology African American ideology Latino ideology
—
—
White Ideology r African American/ Latino Ideology r
OLS
OLS
(2)
Estimator
(1)
Whites and African Americans
1.18** [.06] –4.66** [.90]
—
1.27** [.29] —
—
—
OLS
(6)
.67* [.32] 1.25** [.07] –2.73** [.98]
—
—
—
—
OLS
(7)
1.67* [.74] 1.19** [.07] –5.77* [2.25]
—
—
.44
—
Errors in Vars
(8)
1.16** [.30] —
—
—
OLS
(9)
.36 [.29] 1.17** [.06] –5.44** [1.09]
Whites and Latinos
Table 5.1. Group Differences in Responsiveness, 107th Senate w-nominate Scores
.98 [.79] 1.14** [.07] –6.96** [1.88]
1.05** [.38] —
.44
.95
Errors in Vars
(10)
86 | chapter five
African American ideology drops sharply and becomes statistically insignificant. Accounting for measurement error in the final model of whites and African Americans does not alter the results significantly (see column 5). In sum, these models indicate that, across the full spectrum of issues captured in the w-nominate scores, senators are much less responsive to African Americans’ preferences.12 Indeed, after we account for whites’ preferences, senators appear not to be responsive to the preferences of African Americans at all, which violates any standard of political equality. Turning to a comparison of senators’ responsiveness to whites and Latinos, we find a similar pattern of results (see the right side of table 5.1). We first observe that when we include these measures in the model independently, and before we account for measurement error, senators are more responsive to whites’ preferences than Latinos’ preferences (compare the parameter estimate of 1.27 for white ideology in column 6 to the estimate of 0.67 for Latinos in column 7). Once again, when we correct for measurement error in Latinos’ preferences but not in whites’ preferences, senators appear to be more responsive to Latinos than whites (see column 8). Since Latino ideology is modestly correlated with white ideology, the relationship between Latino ideology and senators’ roll-call votes may be spurious.13 When we include both groups in the model in column 9, the 1.16 estimate for whites far exceeds that for Latinos, and the 0.36 estimate for Latinos in column 9 is hard to distinguish from zero. Finally, when we correct for measurement error in both whites’ and Latinos’ preferences, the estimates for the groups are similar in magnitude, but only whites’ preferences systematically predict senators’ voting decisions at anywhere near traditional levels of statistical significance (see column 10). To illustrate the substantive magnitude of the relationship between constituents’ ideologies and their senators’ votes, we simulated the predicted w-nominate score of senators representing states at the 25th and 75th percentiles of the white and African American and white and to African Americans within states, senators across the country become more conservative in their roll-call behavior (p 90th percentile) are about 0.07 points closer to their MCs than are low-income earners (< 25th percentile). Older Americans are about 0.08 points closer to their
figure 5 .5 . Size of the proximity gap, various groups (2000 National Annenberg Election Survey; Lewis and Poole 2004).
differences in legislative representation | 95
MCs than are younger Americans. Thus, the proximity gap between African Americans and whites represented by non–African American MCs is on par with or even greater than that associated with many of the significant demographic divides that observers have noted in American politics (Bartels 2008 [income]; Campbell 2003 [age]; Griffin, Newman, and Wolbrecht 2006 [gender]). Not only do these gaps provide a point of comparison for the gaps between racial and ethnic groups, but finding that the wealthy and the educated are better represented than the poor and the uneducated, which is in line with prior studies (Gilens 2005; Bartels 2008), helps to validate the proximity approach. Turning to a comparison of whites and Latinos represented by whites and African Americans (columns with diagonals in fig. 5.4), according to all four measures, on average Latinos are farther from their MCs (three of the four measures at p < .01). According to some measures the proximity gap between Latinos and whites represented by non-Latino MCs appears to exceed the gap between African Americans and whites. In light of the diversity of Latino opinion by national origin and (relatedly) by region, we also examined the relative representation of Latino subgroups, finding some important variation. Latinos of Puerto Rican and especially Mexican origin are farther than whites from their MCs (0.04 and 0.13 points farther, respectively).26 In contrast, Latinos of Cuban descent are actually 0.16 points closer than whites on average, though we hesitate to draw inferences based on the small Cuban sample (119 respondents). To the extent that this finding holds, we suspect that this is attributable to the rather homogenous ideological conservatism of Cuban-Americans, a group that is usually represented by Republicans. Regionally, Latinos in the Pacific Coast and Mountain states and Latinos in the Southeast and South Central states show the largest proximity gap compared to whites (0.09 and 0.10 respectively), while the gap is somewhat smaller in the Northeast (0.06). A Closer Look at Race-Conscious Egalitarianism | Thus far we have seen that as groups, whites find their preferences better reflected in their MCs’ voting records than do African Americans or Latinos. However, as we noted earlier, a more widely accepted test of the race-conscious egalitarian standard is to
26. When we control for turnout, income, district racial/ethnic composition, and MC ethnicity and party affiliation, the only substantial change in the results is that Latinos of Puerto Rican origin are no longer significantly further than whites from their MCs.
96 | chapter five
ask whether minorities are just as close as whites to their MCs when these groups constitute a large proportion but not a majority of the constituency. As we discussed in chapter 2, advocates of the race-conscious egalitarian standard would usually require that minorities’ likelihood of being on the winning side of outcomes would be “more than proportional” rather than fully equal. Thus, minorities should not necessarily be represented as well as whites when minorities make up some small share of a constituency, but in districts with substantial minority populations, perhaps greater than 25% or even 40%, more than proportional representation begins to approach a requirement of equal representation. Figure 5.6 compares the relative proximity of minorities and whites across districts with different racial and ethnic compositions, using the Wright measure of ideological proximity. The figure points to two significant findings. First, African Americans remain farther than whites from their MCs even in districts where African Americans make up more than 25%, 40%, or even 50% of a constituency. In fact, if we pool the districts where African Americans make up more than 40% of the constituency, the proximity gap (0.07 points) is about the same as it is in the nation as a whole; however, because there is relatively little data, we cannot be very certain that the mean distances are in this case actually different from each other (p = .19).
figure 5 .6 . Nondescriptive MCs and the proximity gap in the 107th House, by district percentage African American or Latino (2000 National Annenberg Election Survey; Lewis and Poole 2004).
differences in legislative representation | 97
Second, the figure provides some evidence of what some scholars have called “white backlash” against African Americans in districts where they constitute a significant minority of the population (e.g., Key 1949; Blalock 1967; Bullock and Rodgers 1976; Giles and Buckner 1993; but see Lublin 1997). These studies argue that in districts in which African Americans pose an electoral “threat” to whites, whites become politically mobilized, and candidates and elected officials tend to cater to the majority whites’ preferences. The figure shows that the proximity gap between African Americans and whites is greatest in districts where African Americans constitute between 25% and 40% of the population, reaching 0.24 points, more than three times the gap between whites and African Americans in the country as a whole. The results for Latinos are even more telling. Again we see evidence of backlash, as the proximity gap is enormous in districts where Latinos make up a large minority of the constituency. In districts where Latinos comprise 40–50% of the population, Latinos are more than 0.38 points farther than whites from their MCs.27 If we pool the data across multiple categories, in districts where Latinos comprise at least a quarter of the population, Latinos fare worse than they do in all districts (a 0.14 point gap vs. a 0.10 gap nationally; see fig. 5.4). In districts where Latinos make up at least 40% of the constituency, the proximity gap (0.22 points) is more than twice as large as it is in the nation as a whole. To restate the most significant point for our present purposes, figure 5.6 suggests that minorities are not more than proportionally represented—because even when they make up a near-majority of a district’s population, they are not equally represented—and thus that the race-conscious egalitarian standard of equality is not satisfied. We discuss the results for districts that are majority African American or Latino below.
27. Although we cannot be certain, one reason why “backlash” occurs most in districts where African Americans comprise between 25% and 40% of the population and where Latinos comprise a higher percentage (between 40% and 50%) relates to Latinos’ lower rates of citizenship. If the Latino population makes up 40% of a district’s population, some in this group will be ineligible to vote, so the voting-eligible population is actually in the 25–40% range. However, this hypothesis requires whites to be sophisticated enough to estimate the size of the voting-eligible Latino population, as opposed to the entire Latino population. This may be an unreasonable assumption. Another potential reason is more baldly racial in nature—whites may simply consider African Americans more “threatening” than Latinos. At this point, we have no evidence to support either hypothesis.
98 | chapter five
Before we move on, we note that our key finding—that African Americans and Latinos tend to be farther than whites from their MCs—is robust in a number of ways. First, we know that politics is often different in the South, especially when it comes to conflict between whites and African Americans (Key 1949). Racial politics in the South have been more intense and divisive than in other regions. Therefore, we examined whether African Americans are farther away from their MCs than whites in both the South and other areas of the nation. The results show that they are, although African Americans are generally worse off in the South. One reason that they may be worse off in the South is that African Americans’ and whites’ attitudes are more divergent in this region than in other parts of the country.28 If MCs generally side with whites, then the farther African Americans are from whites, the worse off they will be. In addition, given that African Americans generally make up a larger percentage of congressional districts in the South than elsewhere, these results are an early warning that the proportional standard of political equality may not be met for African Americans. We examine this question more directly below. Second, using a regression framework, we controlled for three factors that are likely related to political equality: the racial or ethnic composition of districts (the percentage of each district composed of African Americans or Latinos), constituents’ household income, and self-reported turnout in the 1996 election. Even when controlling for these factors, African Americans and Latinos remained significantly farther from their white MCs than were white constituents. Third, perhaps whites are nearer than African Americans and Latinos to their white MCs because whites are more likely to identify with the same party as their MCs and co-partisans are generally thought to be better represented (Bullock and Brady 1983). However, we found that even among copartisans, African American and Latino constituents were significantly farther from their MCs than were whites (see table 5.2). African American copartisans were an estimated 0.20 points farther than whites from their MCs (see column 1). This gap shrank a bit when we included controls for income and turnout, but remained significant (see column 2). Latino co-partisans were about 0.24 points farther than whites from their MCs, a gap that shrank only marginally when we added controls (see columns 3 and 4).29 28. The difference in the mean ideology (1–5-point scale) of whites and African Americans outside the South is 0.19; in the South, the difference is 0.24. 29. Note that these proximity gaps are actually greater among co-partisans than they are for all constituents (0.20 for co-partisan African Americans, compared to around
differences in legislative representation | 99
Table 5.2. Models of Distance from White MCs among Co-Partisan Citizens and Citizens Represented by Freshmen MCs Citizens represented by freshmen MCs
Co-partisan citizens African American
.199** .142** [.018] [.025]
—
—
.135** [.444]
.166** [.060]
—
—
.236** .195** [.763] [.025]
—
—
.075* [.041]
.025 [.057]
Latino
—
—
Household income
—
–.022** [.003]
—
–.020** [.003]
—
.008 [.007]
—
.008 [.007]
Turnout
—
–.058** [.016]
—
–.072** [.016]
—
–.064* [.035]
—
–.093** [.033]
Constant
.763** .930** .763** .933** 1.01** [.005] [.021] [.005] [.021] [.011]
1.13** [.046]
1.108** 1.144** [.011] [.043]
N
17,505
2,964
6,157
9,436
17,995
9,670
5,824
3,113
Sources: 2000 National Annenberg Election Survey; Lewis and Poole 2004. Note: Dependent variable: Senators’ w-nominate scores, 107th Senate. * p < .05; ** p < .01; standard errors in brackets.
Fourth, if white constituents change their views to match those of their representatives more than other groups do, our evidence could again be misleading. It could be that whites’ preferences are more responsive to the voting behavior of their MCs than are those of African Americans or Latinos when both groups are represented by whites. To limit the possibility of endogeneity, we analyzed the subset of NAES respondents who were represented by a freshman MC in the 107th Congress. This means we used constituent opinion measured before MCs cast a single roll-call vote to model votes from MCs’ first two years in the chamber. This design purges the model of reciprocal effects because MCs’ first-term votes cannot affect their constituents’ ideologies prior to their election. We found that even among citizens who were represented by a freshman member of the House in the 107th Congress,
0.10 for all African Americans; see fig. 5.4). This does not mean that Democratic MCs are farther than Republican MCs from African American or Latino constituents. As we note below, Democrats are much closer than Republicans to their minority constituents. Instead, this result suggests that whites benefit even more than African Americans or Latinos from electing co-partisans.
100 | chapter five
African Americans and (less so) Latinos were politically unequal compared to whites when both groups were represented by whites (see table 5.2). Fifth, some have suggested that MCs may vote more conservatively or liberally on average than their constituents would like (e.g., Bafumi and Herron 2007). That is, members’ votes may generally be more conservative than the w-nominate scores we use would indicate. Therefore, we examined whether shifting MCs as a group in a conservative (or liberal) direction would alter the nature of our conclusions (see fig. 5.7). If we uniformly shift the ideological location of all MCs 0.5 units to the right on the –1 to +1 scale, the results are virtually unchanged for African Americans compared to those we report—they remained further than whites from MCs in general as well as when the African American proportion of the district exceeded 12%, 25%, and even 40% (all p < .05, results not shown). For Latinos, shifting MCs to the right improves their relative representation, both in general and when Latinos make up a larger proportion of districts. For instance, across all districts Latinos are on average 0.07 points further than whites from their MCs (p < .01), whereas this group averaged 0.11 points further prior to the rightward shift. Although this is a better situation for Latinos, note that Latinos continue to be significantly farther from their MCs than are whites. When we shift MCs to the left, the pattern for the groups is reversed—the picture becomes rosier for African Americans, relative to whites, but at least
figure 5 .7 . The proximity gap if Congress is more conservative, liberal, or varied (2000 National Annenberg Election Survey; Lewis and Poole 2004).
differences in legislative representation | 101
as bad if not worse for Latinos compared to whites. Specifically, we find that all African Americans are about 0.055 points farther than whites from their MCs, that there is a smaller backlash in districts with somewhat large African American populations, and that African Americans are no further than whites from their MCs once the African American population tops 40%. All Latinos are 0.12 points further than whites from their MCs, a gap that stays the same as the proportion of the district that is Latino increases to at least 12%, and then grows as this proportion increases to 25% (0.16), and 40% (0.21). We also shifted Democratic MCs to the left and Republican MCs to the right. Upon doing so, we found that the proximity gap between whites and all African Americans remained about the same as it was prior to any manipulation of MCs’ ideological locations (0.086), and this gap widened as the African American population increased, until the population exceeded 40%, at which point the two groups were equally distant from their MCs. Meanwhile the proximity gap between whites and all Latinos was similar to what we originally found (0.11), and this gap again grows as the Latino proportion of the district increases to more than 25% (0.16) or 40% (0.25). In sum, whether MCs as a group are actually more liberal or conservative than our measure suggests, or whether this differs by MC party, African Americans and Latinos remain further than whites from their MCs; systematic shifts in the location of MCs only affect the degree to which groups are unequally represented. Next, it may be that it is difficult for MCs to represent many African Americans and Latinos well simply because they are more likely to reside in districts that are difficult to represent. If a district is heterogeneous, meaning the constituents in that district hold a wide variety of opinions, it will be impossible for any MC to represent this diversity when representation is conceived as proximity. In contrast, in a district in which the overwhelming majority of constituents is liberal or conservative, it is easier for an MC to represent most constituents well. To see whether this might be driving our results, we modeled constituents’ distance from their MCs, controlling for district heterogeneity, household income, and MC party. To measure heterogeneity, following Sullivan (1973) and using data from the census, we calculate the fractionalization of the district’s constituency based on race, income, and educational attainment.30 Just as we anticipated, we find that
30. We calculate this by summing the squared proportions of residents in each category of a characteristic, dividing this sum by the total number of characteristics, and subtracting this value from 1. In calculating this measure, we use five racial subdivisions, three income subdivisions (to increase variation), and seven education subdivisions.
102 | chapter five
district heterogeneity increases the distance of MCs from their constituents (see table 5.3). In addition, and more important for our purposes, we find that African Americans and Latinos are significantly farther than whites from their MCs even after controlling for district heterogeneity. Our final look at proximity gaps to evaluate race-conscious egalitarianism in the House looks at whites’ and African Americans’ proximity to their MCs on a specific issue—environmental protection—rather than overall ideology. We examine environmental protection in relation only to whites and African Americans because, as we saw in chapter 3, these groups’ opinions differ on it (see fig. 3.3), an independent interest group has identified significant roll-call votes in this issue domain, and the issue is more salient for whites than African Americans (see figs. 3.6 and 3.8). As a result of this last point, our theoretical expectation is that whites would be advantaged in this specific domain. However, as we saw in the last chapter, whites were only slightly more often winners on environmental spending decisions compared to African Americans. Here, we extend the analysis of environmental policies beyond budgetary considerations to other votes on environmental issues. Finally, these comparisons afford
Table 5.3. Group Differences in Proximity, Controlling for District Heterogeneity Whites and African Americans African American
Whites and Latinos
0.063** [.014]
Latino
0.074** [0.014]
District heterogeneity
0.326** [0.073]
0.453** [0.068]
Household income
–0.011** [0.002]
–0.011** [0.002]
MC Republican
0.066** [0.007]
0.045** [0.007]
Constant
0.901** 0.040
0.849** [0.038]
N
46,714
50,008
Sources: 2000 National Annenberg Election Survey; Lewis and Poole 2004; U.S. Census. Note: Dependent variable: Distance from MC (Wright measure of distance). * p < .05; ** p < .01; standard errors in brackets.
differences in legislative representation | 103
us an opportunity to extend the broad-brush ideological analyses of this chapter to a specific issue. To compare citizens’ positions on environmental policy to legislators’ actions, we examined key environmental votes identified by the League of Conservation Voters (LCV) in the 107th Congress. We use citizens’ expressed opinions regarding environmental policy to assign them LCV scores, after reflecting the scaled 0–100 scores of representatives so that higher scores indicate less support for the LCV position.31 We then used most of the techniques we have introduced to generate proximity measures.32 These comparisons, summarized in figure 5.8, reveal that, according to two of the three available measures on this issue, whites are closer to their representatives than are African Americans. These results give us greater confidence that whites are more likely than African Americans to get what they want in the domain of environmental policy. In sum, our finding that African Americans and Latinos are farther from their MCs than are whites is robust to a number of related factors. Race-Conscious Egalitarianism and Proximity in the Senate | We also apply the proximity approach to the Senate using the Wright measure of distance. Because each of the NAES respondents is represented by two senators, we compare each NAES respondent’s ideological preference to both of their senators’ voting behavior (again w-nominate in the 107th Congress). These results, reported in figure 5.9, first show that, regardless of the proportion of a state’s population that is African American or Latino, minorities are on average farther than whites from their senators. Even when African Americans or Latinos make up at least a quarter of the state population, they remain significantly farther than whites from their senators’ voting patterns. Second, the proximity gap appears to be generally larger for African Americans than for Latinos. For example, in states where African Americans make up 12% or less of the population, the gap between whites’ and African Americans’ proximity to their senators (0.12 points) is twice the gap between whites and 31. The 2000 NAES question wording was, “Protecting the environment and natural resources—should the federal government do more about this, the same as now, less, or nothing at all?” We assigned citizens who indicated they preferred to increase federal effort to protect the environment a score of 0, those who wanted the same level of support a score of 25, those who desired less support 50, and those who desired no support a score of 100. 32. We could not use the binned technique because the NES did not ask citizens to place their representatives on an environmental scale.
104 | chapter five
figure 5.8. Nondescriptive MCs and the proximity gap on LCV votes, 107th Congress (2000 National Annenberg Election Survey; Web site of the League of Conservation Voters: http://www.lcv.org/scorecard/past-scorecards/ at 2001 and 2002).
figure 5 .9 . The proximity gap in the Senate, by state percentage African American or Latino (2000 National Annenberg Election Survey; Lewis and Poole 2004).
differences in legislative representation | 105
Latinos in states with 12% or fewer Latinos (0.06 points). In states where African Americans comprise at least a quarter of the population, the proximity gap is greater than 0.07 points, but the gap between whites and Latinos is just over 0.04 points in states where Latinos make up at least a quarter of the population. Third, the relative representation of African Americans does not steadily improve where they make up a larger share of the state’s population, although this is generally true for Latinos. One additional advantage of exploring proximity in the Senate is that it allows us to do a limited number of comparisons between specific senators and their constituents. That is, the sample sizes enable us to do some within-state comparisons. Our general approach is to measure each respondent’s distance from his or her MC and then average across a number of different districts. To probe whether this averaging creates unexpected results, we compare the distance between African Americans in California and the California senators and the distance between whites in California to those same senators. Our data provide a large enough sample of African Americans and Latinos to perform meaningful analyses in California, New York, and Texas (for Latinos only). Figure 5.10 presents the proximity gaps
figure 5 .1 0 . The within-state proximity gap (2000 National Annenberg Election Survey; Lewis and Poole 2004).
106 | chapter five
in these states. In each case, the senators were much farther from their co-partisan minority constituents than co-partisan whites. This gives us confidence that our methods of averaging across districts and states are not unduly skewing our results. To summarize our findings to this point, senators’ voting behavior is much more responsive and proximate to whites’ preferences than those of African Americans or Latinos. Indeed, the responsiveness approach indicates that once we control for whites’ preferences, the attitudes of racial and ethnic minorities appear to have no impact at all on the responsiveness of senators’ roll-call patterns. Applying a race-conscious egalitarian standard to this finding, we see that minorities are politically unequal, since even relatively small racial and ethnic groups should be represented to some degree. In addition, the proximity gaps between minorities and whites exceed the gaps related to sex, age, education, and income. Moreover, African Americans and Latinos remain farther than whites from their MCs even where these groups make up a significant share (25–40%) of a district, which also points to a failure of the race-conscious egalitarian standard.
proportionality This chapter’s results thus far may seem somewhat unsurprising, given African Americans’ and Latinos’ generally much smaller share of state and congressional district populations compared to that of whites. Some might argue that if African Americans and Latinos comprise on average 12–14% of state and district populations, each group should exert about that percentage of the total influence on representatives. In this section we test this, examining whether the voting patterns of senators and representatives satisfy the requirements of the proportional standard of political equality. Specifically, we ask if the differences in responsiveness and proximity we report in table 5.1 and figure 5.4 are diminished in states and congressional districts with larger African American and Latino populations. We also see if representatives appear to be more responsive to individual whites than to individual members of minority groups. Proportionality and Responsiveness | To assess whether African Americans and Latinos are proportionally represented in states with larger African American and Latino populations within a responsiveness framework, we first interact African American and Latino ideology with the percentage of
differences in legislative representation | 107
Table 5.4. State Racial/Ethnic Composition and Responsiveness, 107th Senate Whites and African Americans
Whites and Latinos
African American opinion* state percentage African American
–3.209 [8.115]
—
African American opinion
–0.434 [1.307]
—
State percentage African American
6.768 [24.710]
—
Latino opinion* state percentage Latino
—
–.625 [7.442]
Latino opinion
—
1.198 [1.177]
State percentage Latino
—
2.102 [22.707]
White opinion Constant N
4.626** [0.650]
2.946** [.909]
–13.236** [4.133]
–13.164** [4.194]
52
42
Sources: 2000 National Annenberg Election Survey; Lewis and Poole 2004. Note: Dependent variable: Senators’ w-nominate scores. * p < .05; ** p < .01; standard errors in brackets.
each state’s population comprised of African Americans or Latinos as measured by the U.S. Census. If senators respond to their minority constituents more when they constitute a larger part of their constituency, this interaction term will be positive and statistically significant. As table 5.4 shows, the interaction term is negative and fails to reach statistical significance, meaning senators’ relative responsiveness to African Americans and Latinos does not increase in states with larger African American or Latino populations.33
33. Excluding the white ideology term from these models does not alter these results in any important way. We also used a quadratic, linear in the terms model, but did not find evidence of a systematic but nonlinear relationship. Finally, we estimated a parallel errors-in-variables model of the fifteen states where African Americans comprise at
108 | chapter five
The most likely explanation for this, in our view, is that the size of African American and Latino populations in most states is not large enough to boost minorities’ likelihood of winning in the Senate. Indeed, prior studies of the House of Representatives find that, where the size of a district’s African American population is larger, there is not a liberalizing effect on representative voting until African Americans comprise about 40% of the district, a level only a few states approach (Cameron, Epstein, and O’Halloran 1996; Lublin 1997). According to the 2000 Census, the states with the largest concentrations of African Americans, Mississippi and Louisiana, have populations that are 36% and 33% African American, respectively. In New Mexico, 42% of the residents are Latino, as are about 32% of the residents of California and Texas.34 One way to test this explanation is to examine the proportionality standard of equality in the House of Representatives, where African Americans and Latinos, independently and in combination, often comprise a much larger share of electoral district populations. We turn to this in the next section. Some might contend that this evidence is merely inferential, because we have simply found that in districts or states with larger minority populations, minority groups’ representation relative to whites does not improve. This does not necessarily imply that individual African Americans or Latinos are less well represented than whites. There would be no need for a minority group to be better represented when its share of the population is larger if it was ex ante better represented than its numbers merited. Although our evidence thus far casts considerable doubt on this interpretation, we want to test the proportionality standard in as many ways as possible. Therefore we also assess the proportionality standard by examining the responsiveness of House representatives as a group to a national sample of individuals representing racial and ethnic groups.35 For instance, to assess the responsiveness of MCs’ votes to whites’ and African Americans’ preferences, we model each MC’s w-nominate score as a function of his least 12% of the population. This estimation also showed that only white ideology has a positive effect on senators’ general roll-call patterns. From any angle, state racial composition does little to improve responsiveness to minority preferences in the Senate. 34. We also examined an interaction between African American or Latino opinion and the combined percentage of African American and Latino state populations. These interactions were statistically insignificant for both Latinos and African Americans. 35. Our analyses focus on members of the House rather than senators for methodological reasons. Specifically, because the model requires the roll-call data to be replicated for each legislator’s constituents, in order to obtain accurate standard errors we need
differences in legislative representation | 109
Table 5.5. Individual-Level Models of Responsiveness Whites and African Americans
Whites and Latinos
White ideology
.017** [.003]
.018** [.003]
African American ideology
.007** [.002]
—
—
.007** [.003]
.016 [.015]
—
—
.023** [.013]
MC Republican
.866** [.019]
.865** [.019]
N
10,796
10,496
Latino ideology African American Latino
Sources: 2000 National Annenberg Election Survey; Lewis and Poole 2004. Notes: Dependent variable: MCs’ w-nominate scores. The probability that the white ideology and African American ideology estimates are equal is